The Promise of Seamless Cross-Language Communication

I am very interested in text-to-speech, speech-to-text, speech-to-speech (one language to another) and I follow the Whisper project closely that is the only open source project out of OpenAI. Si when Dr. Yann LeCun recently shared a project called SeamlessExpressive on 𝕏 (formerly Twitter) about speech-to-speech, I wanted to try it out. Here is my video of testing it using the limited demo they had on their site:

I don't speak French so not sure how it came out from a translation and expression point of view but it seems interesting. I tried Spanish as well and seem to work same way. This project called Seamless, developed by Meta AI scientists, enables real-time translation across multiple languages while preserving the emotion and style of the speaker's voice. This technology could dramatically improve communication between people who speak different languages.  The key innovation behind Seamless is that it performs direct speech-to-speech translation rather than breaking the process into separate speech recognition, text translation, and text-to-speech synthesis steps. This unified model is the first of its kind to:

  • Translate directly from speech in one language into another.  
  • Preserve aspects of the speaker's vocal style, like tone, pausing, rhythm, and emotion.
  • Perform streaming translation with low latency, translating speech as it is being spoken rather than waiting for the speaker to finish.

Seamless was created by combining three main components the researchers developed: 

  • SeamlessM4T v2 - An improved foundational translation model covering 100 languages.  
  • SeamlessExpressive - Captures vocal style and prosody features like emotion, pausing, and rhythm.
  • SeamlessStreaming - Enables real-time translation by translating speech incrementally.  

Bringing these pieces together creates a system where a Spanish speaker could speak naturally, conveying emotion through their voice, and the system would immediately output in French or Mandarin while retaining that expressive style. This moves us closer to the kind of seamless, natural translation seen in science fiction.

Overcoming Key Challenges

Creating a system like Seamless required overcoming multiple complex challenges in speech translation:  

Data Scarcity: High-quality translated speech data is scarce, especially for preserving emotion/style. The team developed innovative techniques to create new datasets.  

Multilinguality: Most speech translation research focuses on bilingual systems. Seamless translates among 100+ languages directly without needing to bridge through English.

Unified Models: Prior work relied on cascading separate recognition, translation, and synthesis models. Seamless uses end-to-end speech-to-speech models.  

Evaluation: New metrics were created to evaluate the preservation of vocal style and streaming latency.

The impacts of having effective multilingual speech translation could be immense in a world where language continues to divide people. As one of the researchers explained:

"Giving those with language barriers the ability to communicate in real-time without erasing their individuality could make prosaic activities like ordering food, communicating with a shopkeeper, or scheduling a medical appointment—all of which abilities non-immigrants take for granted—more ordinary."

Amazon AWS Re:Invent 2023 Ushers in a New Era for AWS

AWS recently held its annual re:Invent conference, showcasing exciting new offerings that demonstrate the company's continued leadership in cloud computing and artificial intelligence. This year's event had a strong focus on how AWS is pioneering innovations in generative AI to provide real business value to customers.

CEO Adam Selipsky and VP of Data and AI Swami Sivasubramanian headlined the event, announcing breakthrough capabilities spanning hardware, software, and services that mark an inflection point for leveraging AI. AWS is committed to progressing generative AI from leading-edge technology into an essential driver of productivity and insight across industries.

Highlights from Major Announcements

Here are some of the most notable announcements that give a glimpse into the cutting-edge of what AWS is building:

  • Amazon Q - A new AI-powered assistant designed for workplace collaboration that can generate content and code to boost team productivity.  
  • AWS Graviton4 and Trainium2 Chips – The latest generation AWS processor and accelerator chips engineered to enable heavy AI workloads like training and inference.  
  • Amazon Bedrock Expansion – New options to deploy and run custom models and automate AI workflows to simplify integration.
  • Amazon SageMaker Updates – Enhanced capabilities for novices and experts alike to build, train, tune and run machine learning models faster. 
  • Amazon Connect + Amazon Q - Combining AI assistance and customer service software to help agents respond to customers more effectively.

AWS underscored its commitment towards an intelligent future with previews showcasing bleeding edge innovation. This vision crystallizes how human-AI collaboration can transform customer experiences and business outcomes when generative AI becomes an integral part of solution stacks. Re:Invent 2023 ushered in this emerging era.

As the curtain falls on AWS re:Invent 2023, the message is clear: AWS is not just keeping up with the pace of technological evolution; it is setting it. Each announcement and innovation revealed at the event is a testament to AWS's unwavering commitment to shaping a future where technology is not just a tool but a catalyst for unimaginable growth and progress. The journey of AWS re:Invent 2023 is not just about celebrating achievements; it's about envisioning and building a future that's brighter, faster, and more connected than ever before.

Celebrating a Powerhouse of AI: FAIR's First Decade

Today marks an important milestone for Meta's Fundamental AI Research (FAIR) team – 10 years of spearheading advancements in artificial intelligence. When FAIR first launched under the leadership of VP and Chief AI Scientist Yann LeCun in 2013, the field of AI was finding its way. He assembled a team of some of the keenest minds at the time to take on fundamental problems in the burgeoning domain of deep learning. Step by step, breakthrough upon breakthrough, FAIR's collective brilliance has expanded the horizons of what machines can perceive, reason, and generate.

The strides over a decade are simply striking. In object detection alone, we've gone from recognizing thousands of objects to real-time detection, instance segmentation, and even segmenting anything. FAIR's contributions in machine translation are similarly trailblazing – from pioneering unsupervised translation across 100 languages to the recent "No Language Left Behind" feat. 

And the momentum continues unabated. This year has been a standout for FAIR in research impact, with award-garnering innovations across subareas of AI. Groundbreaking new models like Llama are now publicly available—and FAIR's advancements already power products millions use globally.

While future progress will likely come from fusion rather than specialization, one thing is evident – FAIR remains peerless in its ability to solve AI's toughest challenges. With visionary researchers, a culture of openness, and the latitude to explore, they have their sights firmly fixed on the future.

So, to all those who contributed to this decade of ingenuity – congratulations. And here's to many more brilliant, accountable steps in unleashing AI's potential.

The Art of Reading Signals: Making Sense of Intent in the Age of AI

The images that emerged from Cuba in October 1962 shocked the Kennedy administration. Photos from a U-2 spy plane revealed Soviet missile sites under feverish construction just 90 miles off the coast of Florida. The installations posed a direct threat to the U.S. mainland, drastically altering the balance of power that had kept an uneasy peace. In a televised address on October 22, President Kennedy revealed the Soviet deception and announced a blockade to prevent further missiles from reaching Cuba. The world anxiously watched the crisis build over the next tension-filled week. 

Behind the scenes, critical signals were being misread on both sides. Soviet premier Nikita Khrushchev believed the United States knew of Moscow’s inferior strategic position relative to its superpower rival. In secret discussions with Kennedy, Khrushchev voiced dismay that his attempt to redress the imbalance was perceived as offensive rather than a deterrent. Kennedy, blindsided by photographs he never expected to see, questioned why the Soviets would take such a risk over an island nation of questionable strategic value. Faulty assumptions about intent magnified distrust and instability at the highest levels.

The perils of miscommunication that defined the Cuban Missile Crisis feel disturbingly resonant today. Nations compete for advantage in trade, technology, and security matters beyond the horizon of public visibility. Artificial intelligence powers more decisions than ever in governance, finance, transportation, health, and a growing array of sectors. Yet intentions behind rapid AI progress often need to be clarified even between ostensible partners, let alone competitors.   So, how can nations credibly signal intentions around artificial intelligence while managing risks?

The technology and national security policy worlds require prompt solutions - tailor-made connections enabling credible communication of intentions around artificial intelligence between governments, companies, researchers, and public stakeholders. We will explore critical insights from a crucial recent analysis titled “Decoding Intentions: Artificial Intelligence and Costly Signals " to demystify the AI landscape.” by Andrew Imbrie, Owen Daniels, and Helen Toner.  Ms. Toner has recently come to the limelight in the recent OpenAI saga as she is one of the OpenAI Board of Directors who fired Sam Altman, the co-founder and reinstated CEO of OpenAI.

The core idea is that verbal statements or physical actions that impose political, economic, or reputational costs for the signaling nation or group can reveal helpful information about underlying capabilities, interests, incentives, and timelines between rivals. Their essential value and credibility lie in the potential price the sender would pay in various forms if their commitments or threats ultimately went unfulfilled. Such intentionally “costly signals” were critical, if also inevitably imperfect, tools that facilitated vital communication between American and Soviet leaders during the Cold War. This signaling model remains highly relevant in strategically navigating cooperation and competition dynamics surrounding 21st-century technological transformation, including artificial intelligence. The report identifies and defines four mechanisms for imposing costs that allow nations or companies employing them to signal information credibly:

Tying hands rely on public pledges before domestic or international audiences, be they voluntary commitments around privacy or binding legal restrictions mandating transparency. Suppose guarantees made openly to constituents or partners are met down the line. In that case, political leaders can avoid losing future elections, or firms may contend with angry users abandoning their platforms and services. Both scenarios exemplify the political and economic costs of reneging on promises. 

Sunk costs center on significant one-time investments or resource allocations that cannot be fully recovered once expended. Governments steering funds toward research on AI safety techniques or companies dedicating large budgets for testing dangerous model behaviors signal long-standing directional buy-in. 

Installment costs entail incremental future payments or concessions instead of upfront costs. For instance, governments could agree to allow outside monitors regular and sustained access to continually verify properties of algorithmic systems already deployed and check that they still operate safely and as legally intended. 

Reducible costs differ by being paid mainly at the outset but with the potential to be partially offset over an extended period. Firms may invest heavily in producing tools that increase algorithmic model interpretability and transparency for users, allowing them to regain trust - and market share - via a demonstrated commitment to responsible innovation.

In assessing applications of these signaling logics, the analysis spotlights three illuminating case studies: military AI intentions between major rivals, messaging strains around U.S. promotion of “democratic AI,” and private sector attempts to convey restraint regarding impactful language model releases. Among critical implications, we learn that credibly communicating values or intentions has grown more challenging for several reasons. Signals have become “noisier” overall amid increasingly dispersed loci of innovation across borders and non-governmental actors. Public stands meant to communicate commitments internally may inadvertently introduce tensions with partners who neither share the priorities expressed nor perceive them as applicable. However, calibrated signaling remains a necessary, if frequently messy, practice essential for stability. If policymakers expect to promote norms effectively around pressing technology issues like ubiquitous AI systems, they cannot simply rely upon the concealment of development activities or capabilities between competitors.

Rather than a constraint, complexity creates chances for tailoring solutions. Political and industry leaders must actively work to send appropriate signals through trusted diplomatic, military-to-military, scientific, or corporate channels to reach their intended audiences. Even flawed messaging that clarifies assumptions reassures observers, or binds hands carries value. It may aid comprehension, avoid misunderstandings that spark crises or embed precedents encouraging responsible innovation mandates more widely. To this end, cooperative multilateral initiatives laying ground rules around priorities like safety, transparency, and oversight constitute potent signals promoting favorable norms. They would help democratize AI access and stewardship for the public good rather than solely for competitive advantage. 

When American and Soviet leaders secretly negotiated an end to the Cuban Missile Crisis, both sides recognized the urgent necessity of installing direct communication links and concrete verification measures, allowing them to signal rapidly during future tensions. Policymakers today should draw wisdom from this model and begin building diverse pathways for credible signaling right now before destabilizing accidents occur, not during crisis aftermaths. Reading accurate intent at scale will remain an art more than deterministic science for the foreseeable future.

The Two Minds: How Humans Think Smarter Than AI

I follow Dr. Yann LeCun on 𝕏 (formerly Twitter) as he engages the public on AI's complex science and ethics. His involvement gives me hope the best minds work toward beneficial AI. Recently, he engaged in a Twitter discourse that prompted me to write this post. 

Dr. LeCun has been very clear about the limitations of the Large Language Model (LLM) for a long time. Sadly,  a good chunk of the social media crowd freaks out about how close we are to Artificial General Intelligence (AGI), human-level intelligence. They come to this conclusion based on their interactions with LLMs, which are very effective role-playing and token prediction engines trained on the written text of modern humans. 

Dr. LeCun argues that even the mightiest AI needs more human/animal reasoning and planning. Where does this gap arise from? Dr. LeCun highlights fast, instinctive thought versus deliberate analysis.  

In "Thinking Fast and Slow," Daniel Kahneman described System 1 for instinctive reaction and System 2 for deeper consideration, enabling complex planning.  

Today's AI uses reactive System 1, thinking like a baseball player effortlessly swinging. Per Dr. LeCun, "LLMs produce answers with fixed computation–no way to devote more effort to hard problems." While GPT-3 responds fluidly, it cannot iterate toward better solutions using causality models, the essence of System 2.

Systems like the chess engine AlphaZero better showcase System 2 reasoning by uncovering counterintuitive long-term plans after learning the game's intricacies. Yet modeling cause-and-effect in the real world still challenges AI, according to Dr. LeCun. 

Dr. LeCun argues that planning AI needs "world models" to forecast the outcomes of different action sequences. However, constructing sufficiently advanced simulations remains an open problem. Dr. LeCun notes that "hierarchical planning" compounding objectives still eludes AI while easy for humans/animals. Mastering planning requires "world models" that extrapolate decisions' cascading impacts over time and space like humans intuitively can.

Meanwhile, raw reasoning alone cannot replicate human intelligence. Equally crucial is common sense from experiencing the natural world's messy complexity. This likely explains AI's glaring oversights about ordinary events compared to humans. Combining reasoning prowess with curated knowledge of how the world works offers exciting possibilities for AI with balanced reactive and deliberative capacities akin to our minds.

The great paradox of AI is that models can far exceed humans at specific tasks thanks to computing, yet lacking general thinking skills. Closing this reasoning gap is essential for robust, trustworthy AI. Dr. LeCun's insights guide integrating planning, causality, common sense, and compositionality into AI. Doing so inches us closer to artificial general intelligence that lives up to its name.

Want to Follow Dr. LeCun and Other Top Scientists?

Like Dr. LeCun, CPROMPT.AI tracks 130+ top AI scientists by checking their social media profiles updating their information in a single directory of WHO IS WHO of AI. 



How AI is Transforming Software Development

Enter Poolside AI, an innovative startup founded in 2023 by Jason Warner and Eiso Kant. Jason Warner, with a background as a VC at Redpoint, former CTO at GitHub, and leader of engineering teams at Heroku and Canonical, brings extensive experience in technology and leadership.

The main goal of Poolside AI is to democratize software development by enabling users to instruct their tools in natural language. This approach makes software development more accessible, allowing even non-coders to create applications. The company is developing a ChatGPT-like AI model for generating software code through natural language, aligning with the broader trend of AI-driven software development​​​​.

The startup, based in the US, raised $126 million in a seed funding round, an extension of an initial $26 million seed round announced earlier. French billionaire Xavier Niel and the US venture capital firm Felicis Ventures led this funding.  Moreover, Poolside AI is focused on pursuing Artificial General Intelligence (AGI) for software creation. This ambitious goal underlines their commitment to unlocking new potentials in the field of software development with the backing of investors like Redpoint.

We have grown accustomed to thinking of coding as an elite skill - the specialized domain of software engineers and computer scientists. For decades, the ability to write software has been seen as an esoteric, even mystical, capability accessible only to those willing to devote years to mastering abstract programming languages like C++, Java, and Python. That is beginning to change in profound ways that may forever alter the nature of software creation.

The recent explosion of AI technologies like ChatGPT and GitHub Copilot presages a tectonic shift in how we produce the code that runs everything from websites to mobile apps to algorithms trading billions on Wall Street. Instead of endlessly typing lines of code by hand, a new generation of AI agents promises to generate entire programs on demand, converting basic prompts categorized in plain English into robust functioning software in seconds. 

Just ask Alex, a mid-career developer with over ten years of experience building web applications for startups and enterprise companies. He has honed his craft over thousands of late-night coding sessions, poring over logic errors and debugging tricky bits of database code. Now, with the advent of open AI models like Codex and Claude that can churn out passable code from simple descriptive prompts, Alex feels a creeping sense of unease.

In online developer forums Alex haunts, heated arguments have broken out about what AI-generated code means for traditional programmers. The ability of nonexperts to produce working software without traditional skills strikes many as an existential threat. Some insist that truly skilled engineers will always be needed to handle complex programming tasks and make high-level architectural decisions. But others point to AI achievements like DeepMind's AlphaCode outperforming human coders in competitive programming contests as harbingers of automation in the industry.

Having invested so much time mastering his trade, the prospect fills Alex with dread. He can't shake a feeling that software development risks becoming a blue-collar profession, cheapened by AI that floods the market with decent enough code to undercut human programmers. Rather than a meritocracy rewarding analytical ability, career success may soon depend more on soft skills - your effectiveness at interfacing with product managers and designers using AI tools to translate their visions into reality.

The anxiety has left Alex questioning everything. He contemplates ditching coding altogether for a more AI-proof career like law or medicine - or even picking up trade skills as a carpenter or electrician. At a minimum, Alex knows he will have to specialize in some niche software subdomain to retain value. But with two kids and a mortgage, the uncertainty has him losing sleep at night.

Alex's qualms reflect a burgeoning phenomenon I call AI Anxiety Disorder. As breakthroughs in profound learning alchemy increasingly automate white-collar work once thought beyond the reach of software, existential angst is spreading among knowledge workers. Just as blue-collar laborers came to fear robotics eliminating manufacturing jobs in the 20th century, today's programmers, paralegals, radiologists, and quantitative analysts nervously eye advancements in generative AI as threats to their livelihood. 

Symptoms run from mild unease to total-blown panic attacks triggered by news of the latest AI milestone. After all, we have seen technology disrupt entire industries before - digital photography decimating Kodak and Netflix's devastating Blockbuster Video. Is coding next on the chopping block?

While understandable, allowing AI anxiety to fester is counterproductive. Beyond needless stress, it obscures the bigger picture that almost certainly includes abundant coding opportunities on the horizon. We would do well to remember that new technologies enable as much as they erase. The locomotive put blacksmiths out of work but created orders of magnitude more jobs. The proliferation of cheap home PCs extinguished secretaries' careers typing memos but launched a thousand tech startups. 

And early indications suggest AI will expand rather than shrink the need for software engineers. Yes, AI can now spit out simple CRUD apps and scripting glue code. But transforming those narrow capabilities into full-stack business solutions requires humans carefully orchestrating complementary tools. Foreseeable bottlenecks around design, integration, testing, and maintenance ensure coding jobs are around for a while.

But while AI won't wipe out programming jobs, it will markedly change them. Coders in the coming decades can expect to spend less time performing repetitive coding tasks and more time on higher-level strategic work - distilling opaque requirements into clean specifications for AI to implement and ruthlessly evaluating the output for hidden flaws. Successful engineers will combine critical thinking and communication skills to toggle between human and artificial team members seamlessly.

Tomorrow's programmers will be chief conductors of programming orchestras, blending human musicians playing custom instruments and AI composers interpreting the score into harmonious code—engineers who are unwilling or unable to adapt and risk being left behind.

The good news is that early adopters stand to gain the most from AI's rise. While novice coders may increasingly break into the field relying on AI assistance, experts like Alex are best positioned to synthesize creative solutions by leveraging AI. The most brilliant strategy is to intimately learn the capacities and limitations of tools like GitHub Copilot and Claude to supercharge productivity.

AI anxiety stems from understandable instincts. Humanity has long feared our creations exceeding their creators. From Golem legends to Skynet doomsday scenarios, we have worried about being replaced by our inventions. And to be sure, AI will claim some coding occupations previously thought inviolable, just as past breakthroughs rendered time-honored professions obsolete.

But rather than dread the future, forward-looking coders should focus on the plethora of novel opportunities AI will uncover. Automating the tedious will let us concentrate creativity on the inspired. Working symbiotically with artificial allies will generate marvels unimaginable today. AI will only expand the frontier of software innovation for those agile enough to adapt.

The coming changes will prove jarring for many incumbent programmers accustomed to old working methods. However, software development has always demanded learning nimble new languages and environments regularly. AI represents the latest skill to integrate into a modern coder's ever-expanding toolkit.

It is early days, but the robots aren't here to replace the coders. Instead, they have come to code beside us. The question is whether we choose to code with them or sit back and allow ourselves to be coded out of the future.

Effective Altruism Unpacked: Ethics, AI, and Philanthropy

When we think about making the world a better place, most imagine donating to charities that tug at our heartstrings - feeding hungry children, housing people experiencing homelessness, saving endangered animals. These are all worthy causes, but are they the most effective way to help humanity? An effective altruism movement argues that we should decide how to spend our time, money, and energy based on evidence and reason rather than emotion. 

Effective altruists try to answer a simple but surprisingly tricky question - how can we best use our resources to help others? Rather than following our hearts, they argue we should track the data. By taking an almost business-like approach to philanthropy, they aim to maximize the “return” on each dollar and hour donated.

The origins of this movement can be traced back to Oxford philosopher William MacAskill. As a graduate student, MacAskill recognized that some charities manage to save lives at a tiny fraction of the cost of others. For example, the Against Malaria Foundation provides insecticide-treated bed nets to protect people from malaria-carrying mosquitos. This simple intervention costs just a few thousand dollars per life saved. Meanwhile, some research hospitals spend millions of dollars pursuing cutting-edge treatments that may hold only a handful of patients. 

MacAskill realized that a small donation to a highly effective charity could transform many more lives than a large donation to a less efficient cause. He coined effective altruism to describe this approach of directing resources wherever they can have the most significant impact. He began encouraging fellow philosophers to treat charity not as an emotional act but as a mathematical optimization problem - where can each dollar do the most good?

Since its beginnings in Oxford dorm rooms, effective altruism has become an influential cause supported by Silicon Valley tech moguls and Wall Street quants. Figures like Bill Gates, Warren Buffet, and hedge fund manager John Paulson have all incorporated practical altruist principles into their philanthropic efforts. Instead of arbitrarily dividing their charitable budgets between causes that inspire them personally, they rely on research and analysis to guide their giving.

The influential altruism community has influenced how the ultra-rich give and how everyday people spend their time and disposable income. Through online communities and local groups, thousands of professionals connect to discuss which careers and activities could positively impact society. Rather than arbitrarily pursuing work they find interesting, many effective altruists choose career paths specifically intended to do the most good - even if that means working for causes they do not have a passion for.

For example, graduates from top universities are now forgoing high-paying jobs to work at effective charities they have researched and believe in. Some conduct randomized controlled trials to determine which development interventions work so charities can appropriately direct funding. Others analyze the cost-effectiveness of policy changes related to global issues like pandemic preparedness and climate change mitigation. Even those in conventional corporate roles aim to earn higher salaries to donate to thoroughly vetted, effective charities substantially.

However, in recent years, AI safety has emerged as one of the most prominent causes within the influential altruist community - so much so that some now conflate effective altruism with the AI safety movement. This partly stems from Nick Bostrom’s influential book Superintelligence, which made an ethical case for reducing existential risk from advanced AI. Some effective altruists found Bostrom’s argument compelling, given the immense potential consequences AI could have on humanity’s trajectory. The astronomical number of hypothetical future lives that could be affected leads some to prioritize AI safety over more immediate issues. 

However, others criticize this view as overly speculative doom-saying that redirects attention away from current solvable problems. Though they agree advanced AI does pose non-negligible risks, they argue the probability of existential catastrophe is extremely hard to estimate. They accuse the AI safety wing of the movement of arbitrarily throwing around precise-sounding yet unfounded statistics about extinction risks.

Despite these debates surrounding AI, the effective altruism movement continues working to reshape attitudes toward charity using evidence and logical reasoning. Even those skeptical of its recent focus on speculative threats agree the underlying principles are sound - we should try to help others as much as possible, not as much as makes us feel good. By taking a scientific approach to philanthropy, effective altruists offer hope that rational optimism can prevail over emotional pessimism when tackling the world’s problems.

Frequently Asked Questions

Q: How is effective altruism different from other forms of charity or activism?

A: The effective altruism movement emphasizes using evidence and reason to determine which causes and interventions do the most to help others. This impartial, mathematical approach maximizes positive impact rather than supporting causes based on subjective values or emotions.

Q: Who are some notable people associated with effective altruism? 

A: Though it originated in academic philosophy circles at Oxford, effective altruism now encompasses a range of influencers across disciplines. Well-known figures like Bill Gates, Warren Buffet, and Elon Musk have all incorporated practical altruist principles into their philanthropy and business initiatives. 

Q: What are some examples of high-impact career paths effective altruists pursue?

A: Many effective altruists select careers specifically to impact important causes positively. This could involve scientific research on climate change or pandemic preparedness that informs better policy. It also includes cost-effectiveness analysis for charities to help direct funding to save and improve the most lives per dollar. 

Q: Do effective altruists only focus on global poverty and health issues? 

A: While saving and improving lives in the developing world has been a significant focus historically, the movement now spans a wide range of causes. However, debate surrounds whether speculative risks like advanced artificial intelligence should be considered on par with urgent humanitarian issues that could be addressed today.

Q: Is effective altruism relevant to people with little time or money to donate?

A: Yes - effective altruism provides a framework for integrating evidence-based decision-making into everyday choices and habits. Knowing which behaviors and purchases drive the most positive impact can help ordinary people contribute to the greater good through small but systemic lifestyle changes.

Bridging the Gap Between Humans and AI

We stand at a unique moment in history, on the cusp of a technology that promises to transform society as profoundly as the advent of electricity or the internet age. I'm talking about artificial intelligence (AI) - specifically, large language models like ChatGPT that can generate human-like text on demand. 

In a recent conference hosted by the World Science Festival, experts gathered to discuss this fast-emerging field's awe-inspiring potential and sobering implications. While AI's creative capacity may wow audiences, leading minds urge us to peer under the hood and truly understand these systems before deploying them at scale. Here is the video:

The Core Idea: AI is Still Narrow Intelligence

ChatGPT and similar large language models use self-supervised learning on massive datasets to predict text sequences, even answering questions or writing poems. Impressive, yes, but as AI pioneer Yann LeCun cautions, flexibility with language alone does not equate to intelligence. In his words, "these systems are incredibly stupid." Compared to animals, AI cannot perceive or understand the physical world. 

LeCun stresses current AI has no innate desire for domination. Still, it lacks judgment, so safeguards are needed to prevent misuse while allowing innovation for social good. For example, CPROMPT.AI will enable users without coding skills to build and share AI apps quickly and easily, expanding access to technology for a more significant benefit. LeCun's vision is an open-source AI architecture with a planning capacity more akin to human cognition. We have yet to arrive, but steady progress brings this within reach.

Emergent Intelligence

What makes ChatGPT so adept with words? Microsoft's Sebastian Bubeck reveals it's based on a transformer architecture system. This processes sequences (like sentences) by comparing words to other words in context. Adding more and more of these comparison layers enables the identification of elaborate pattern patterns. So, while its world knowledge comes from digesting some trillion-plus words online, the model interrelates concepts on a vast scale no human could match. Still, current AI cannot plan; it can only react.

Can We Control the Trajectory? 

Tristan Harris of the Center for Humane Technology warns that AI applications are already impacting society in unpredictable ways. Their incentives -- engagement, speed, scale -- don't align with human wellbeing. However, Bubeck suggests academic research motivated by understanding, not profit, can point the way. His team created a mini-model that avoids toxic online content. AI could gain beneficial skills without detrimental behaviors with thoughtfully curated data and testing. 

Progress Marches Onward  

"This is really incredible," remarks Bubeck - who never expected to see such advances in his lifetime. Yet he cautions that capacities are compounding at a clip beyond society's adjustment rate. We must guide this technology wisely. What role will each of us play in shaping how AI and humans coexist? We don't have to leave it up to tech titans and policymakers. Every time we use CPROMPT.AI to create an AI-powered app, we direct its impact in a small way. This epoch-defining technology ultimately answers to the aspirations of humanity. Where will we steer it next?


  • Transformer architecture: The system underlying ChatGPT and other large language models, using comparison of words in context to predict patterns
  • Self-supervised learning: Training AI models to perform a task by giving examples rather than explicit rules (e.g., predicting missing words) 
  • CPROMPT.AI: A platform allowing easy no-code creation of AI apps to share 

Q* | OpenAI | 𝕏

Recently, a prominent Silicon Valley drama took place -- the OpenAI CEO, Sam Altman, was fired by his board and rehired after pressure from Microsoft and OpenAI employees. Employees allegedly threatened to leave the company if Altman was not reinstated. Microsoft assisted with handling the crisis and returning Altman to his CEO role.  I won't go into the details of the drama but I will provide you with a summary card below that covers my analysis of this saga.

As this unfolded on Twitter, gossip emerged that a specific OpenAI development had concerned the board. They allegedly believed Altman needed to be more truthful about the state of progress toward AGI (artificial general intelligence) within the company. This led to speculation and conspiracy theories on Twitter, as often happens with high-profile industry drama. 

One theory pointed to OpenAI's advancements with an algorithm called Q*. Some suggested Q* allowed internal LLMs (large language models) to perform basic math, seemingly bringing OpenAI closer to more advanced AI. In this post, I'll explain what Q* is and why its advancements could theoretically bring AI systems closer to goals like AGI.  

What is Q*?

In simple terms, Q* is like a GPS that learns over time. Usually, when there's traffic or an accident, your GPS doesn't know and tries to lead you to the usual route, which gets stuck. So, you wait for it to recalculate a new path fully. What if your GPS started remembering problems and closures so that next time, it already knows alternate routes? That's what Q* does. 

Whenever Q* searches for solutions, like alternate directions, it remembers what it tried before. This guides future searches. So if something changes along a route, Q* doesn't restart like a GPS recalculating. It knows most of the road and can focus only on adjusting the tricky, different parts.  

This reuse makes Q* get answers faster than restarting every time. It "learns" from experience, like you learning backroad ways around town. The more Q* is used, the better it adapts to typical area changes.

Here is a more technical explanation:

Q* is an influential algorithm in AI for search and pathfinding. Q* extends the A* search algorithm. It improves A* by reusing previous search efforts even as the environment changes. This makes it efficient for searches in dynamic environments. Like A*, Q* uses a heuristic function to guide its search toward the goal. It balances exploiting promising areas (the heuristic) with exploring new areas (like breadth-first search). Q* leverages experience from previous searches to create a reusable graph/tree of surveyed states. 

This significantly speeds up future searches rather than starting fresh each time. As the environment changes, Q* updates its reusable structure to reflect changes rather than discarding it. 

This allows reusing valid parts and only researching affected areas. Q* is famously used for robot path planning, manufacturing, and video games where environments frequently change. It allows agents to replan paths as needed efficiently.

In summary, Q* efficiently finds solutions in systems where the state space and operators change over time by reusing experience. It can discover solutions much faster than restarting the search from scratch.

So, in the context of the rumors about OpenAI, some hypothesize that advances leveraging Q* search techniques could allow AI and machine learning models to more rapidly explore complex spaces like mathematics. Rather than re-exploring basic rules from scratch, models might leverage prior search "experience" and heuristics to guide discovery. This could unlock new abilities and general skills.

However, whether OpenAI has made such advances leveraging Q* or algorithms like it is speculative. The details are vague, and rumors should be critically examined before conclusions are drawn. But Q* illustrates interesting AI capabilities applicable in various domains. And it hints at future systems that may learn and adapt more and more like humans.

The Controversial Decision to Award OpenAI the 2023 Hawking Fellowship

The Cambridge Union and Hawking Fellowship committee recently announced their controversial decision to jointly award the 2023 Hawking Fellowship to OpenAI, the creators of ChatGPT and DALL-E. While OpenAI is known for its advancements in AI, the award has sparked debate on whether the company truly embodies the values of the fellowship. 

What the committee saw in OpenAI:

  • OpenAI has successfully shifted perceptions about what AI is capable of through innovations like ChatGPT. Their models represent significant progress in natural language processing.
  • The company has committed to releasing most of its open-source AI work and making products widely accessible. 
  • OpenAI espouses responsible development of AI to benefit humanity, which aligns with the spirit of the Hawking Fellowship.

However, as a well-funded startup, OpenAI operates more like a tech company than an altruistic non-profit acting for the public good. Its mission to create and profit from increasingly capable AI systems takes precedence over caution. There are concerns about the potential dangers of advanced AI systems that could be misused.

Anyway, in case you didn't watch the above video, here is what Sam Altman's speech highlighted:

  • AI has extraordinary potential to improve lives if developed safely and its benefits distributed equitably. 
  • OpenAI aims to create AI that benefits all humanity, avoiding the profit maximization incentives of big tech companies.
  • They are working to develop safeguards and practices to ensure robust AI systems are not misused accidentally or intentionally.
  • Democratizing access to AI models allows more people to benefit from and provide oversight on its development. 
  • OpenAI is committed to value alignment, though defining whose values to align with poses challenges.
  • Another breakthrough beyond improving language models will likely be needed to reach advanced general intelligence.

While OpenAI is making impressive progress in AI, reasonable concerns remain about safety, ethics, and the company's priorities as it rapidly scales its systems. The Hawking Fellowship committee took a gamble in awarding OpenAI, which could pay off if they responsibly deliver on their mission. But only time will tell whether this controversial decision was the right one.


Q: What is OpenAI's corporate structure?

OpenAI started as a non-profit research organization in 2015. In 2019, they created a for-profit entity controlled by the non-profit to secure funding needed to develop advanced AI systems. The for-profit has a capped return for investors, with excess profits returning to the non-profit. 

Q: Why did OpenAI change from a non-profit? 

As a non-profit, OpenAI realized it could need more time, tens or hundreds of billions required to develop advanced AI systems. The for-profit model allows them to access capital while still pursuing their mission.

Q: How does the structure benefit OpenAI's mission?

The capped investor returns and non-profit governance let OpenAI focus on developing AI to benefit humanity rather than pursuing unlimited profits. The structure reinforces an incentive system aligned with their mission.

Q: Does OpenAI retain control of the for-profit entity? 

Yes, the non-profit OpenAI controls the for-profit board and thus governs significant decisions about the development and deployment of AI systems.

Q: How does OpenAI use profits to benefit the public?

As a non-profit, any profits of the for-profit above the capped returns can be used by OpenAI for public benefit. This could include aligning AI with human values, distributing benefits equitably, and preparing society for AI impacts.

Q: What is Sam Altman's perspective on how universities need to adapt to AI?

Sam Altman believes that while specific curriculum content and educational tools will need to adapt to advances in AI, the core value of university education - developing skills like critical thinking, creativity, and learning how to learn across disciplines - will remain unchanged. Students must fully integrate AI technologies to stay competitive, but banning them out of fear would be counterproductive. Educators should focus on cultivating the underlying human capacities that enable transformative thinking, discovery, and problem-solving with whatever new tools emerge. The next generation may leapfrog older ones in productivity aided by AI, but real-world critical thinking abilities will still need honing. Universities need to modernize their mediums and content while staying grounded in developing the fundamental human skills that power innovation.

Q: What did Sam say about British approach to AI?

Sam Altman spoke positively about the emerging British approach to regulating and governing AI, portraying the UK as a model for thoughtful and nuanced policymaking. He admires the sensible balance the UK government is striking between safely oversighting AI systems while still enabling innovation. Altman highlighted the alignment across government, companies, and organizations in acknowledging the need for AI safety precautions and regulation. At the same time, the UK approach aims to avoid reactionary measures like banning AI development altogether. Altman sees excellent promise in constructive dialogues like the UK AI Summit to shape solutions on governing AI responsibly. He contrasted the reasonable, engaged UK approach to more polarized stances in other countries. Altman commended the UK for its leadership in pragmatically debating and formulating policies to ensure AI benefits society while mitigating risks.

Q: What does Sam think are the critical requirements of a startup founder?

Here are five essential requirements Sam Altman discussed for startup founders:

Determination - Persistence through challenges is critical to success as a founder. The willingness to grind over a long period is hugely important.

Long-term conviction - Successful founders deeply believe in their vision and are willing to be misunderstood long before achieving mainstream acceptance. 

Problem obsession - Founders need an intense focus on solving a problem and commitment to keep pushing on it.

Communication abilities - Clear communication is vital for fundraising, recruitment, explaining the mission, and being an influential evangelist for the startup.

Comfort with ambiguity - Founders must operate amidst uncertainty and keep driving forward before formulas or models prove out.

Q: Why does Sam think the computed threshold needs to be high?

Here are the key points on why Sam Altman believes the computed threshold needs to be high for advanced AI systems requiring oversight:

  • Higher computing power is required to train models that reach capabilities, posing serious misuse risks.
  • Lower capability AI systems can provide valuable applications without the exact oversight needs.
  • If the computed threshold is too low, it could constrain beneficial innovation on smaller open-source models.
  • Altman hopes algorithmic progress can keep the dangerous capability threshold high despite hardware advances reducing compute costs.
  • If capabilities emerge at lower compute levels than expected, it would present challenges for governance.
  • But for now, he thinks truly concerning AI abilities will require large-scale models only accessible to significant players.
  • This makes it feasible to regulate and inspect those robust systems above a high compute threshold.
  • Allowing continued open access to lower capability systems balances openness and safety.
  • In summary, a high compute/capability bar enables oversight of risky AI while encouraging innovation on systems not reaching that bar.

Q: How does Sam think value alignment will work for making ethical AI?

Here are the key points on how Sam Altman believes value alignment will allow the development of ethical AI:

  • Part one is solving the technical problem of aligning AI goal systems with human values.
  • Part two is determining whose values should be aligned with - a significant challenge. 
  • Having AI systems speak with many users could help represent collective moral preferences.
  • This collaborative process can define acceptable model behavior and resolve ethical tradeoffs.
  • However, safeguards are needed to prevent replicating biases that disenfranchise minority voices.
  • Global human rights frameworks should inform the integration of values.
  • Education of users on examining their own biases may be needed while eliciting perspectives.
  • The system can evolve as societal values change.
  • Altman believes aligning AI goals with the values of impacted people is an important starting point. 

However, the process must ensure representative input and prevent codifying harmful biases. Ongoing collaboration will be essential.

Q: What does Sam say about the contemporary history of all technologies?

Sam Altman observed that there has been a moral panic regarding the negative consequences of every major new technology throughout history. People have reacted by wanting to ban or constrain these technologies out of fear of their impacts. However, Altman argues that without continued technological progress, the default state is decay in the quality of human life. He believes precedents show that societal structures and safeguards inevitably emerge to allow new technologies to be harnessed for human benefit over time. 

Altman notes that prior generations created innovations, knowing future generations would benefit more from building on them. While acknowledging new technologies can have downsides, he contends the immense potential to improve lives outweighs the risks. Altman argues we must continue pursuing technology for social good while mitigating dangers through solutions crafted via societal consensus. He warns that abandoning innovation altogether due to risks would forego tremendous progress.

Q: What does Sam think about companies that rely on advertising for revenue, such as the social media mega-companies?

Sam Altman said that while not inherently unethical, the advertising-based business model often creates misaligned incentives between companies and users. He argued that when user attention and data become products to be exploited for revenue, it can lead companies down dangerous paths, prioritizing addiction and engagement over user well-being. Altman observed that many social media companies failed to implement adequate safeguards against harms like political radicalization and youth mental health issues that can emerge when systems are designed to maximize engagement above all else. However, he believes advertising-driven models could be made ethical if companies prioritized societal impact over profits. Altman feels AI developers should learn from the mistakes of ad-reliant social media companies by ensuring their systems are aligned to benefit society from the start.

Q: What does Sam think about open-source AI?

Sam Altman said he believes open-sourcing AI models are essential for transparency and democratization but should be done responsibly. He argued that sharing open-source AI has benefits in enabling public oversight and access. However, Altman cautioned that indiscriminately releasing all AI could be reckless, as large models should go through testing and review first to avoid irreversible mistakes. He feels there should be a balanced approach weighing openness and precaution based on an AI system's societal impact. Altman disagrees with both altogether banning open AI and entirely unfettered open sourcing. He believes current large language models are at a scale where open source access makes sense under a thoughtful framework, but more advanced systems will require oversight. Overall, Altman advocates for openness where feasible but in a measured way that manages risks.

Q: What is Sam's definition of consciousness?

When asked by an attendee, Sam Altman did not provide his definition of consciousness but referenced the Oxford Dictionary's "state of being aware of and responsive to one's surroundings." He discussed a hypothetical experiment to detect AI consciousness by training a model without exposure to the concept of consciousness and then seeing if it can understand and describe subjective experience anyway. Altman believes this could indicate a level of consciousness if the AI can discuss the concept without prior knowledge. However, he stated that OpenAI has no systems approaching consciousness and would inform the public if they believe they have achieved it. Overall, while not explicitly defining consciousness, Altman described an experimental approach to evaluating AI systems for potential signs of conscious awareness based on their ability to understand subjective experience despite having no training in the concept.

Q: What does Sam think about energy abundance affecting AI safety?

Sam Altman believes energy abundance leading to cheaper computing costs would not undermine AI safety precautions in the near term but could dramatically reshape the landscape in the long run. He argues that while extremely affordable energy would reduce one limitation on AI capabilities, hardware and chip supply chain constraints will remain bottlenecks for years. However, Altman acknowledges that abundant clean energy could eventually enable the training of models at unprecedented scales and rapidity, significantly accelerating the timeline for advancing AI systems to transformative levels. While he feels risks would still be predictable and manageable, plentiful energy could compress the progress trajectory enough to substantially impact the outlook for controlling super-advanced AI over the long term. In summary, Altman sees energy breakthroughs as not negating safety in the short term but potentially reshaping the advancement curve in the more distant future.