Is 'The AI Revolution" A Scam ?

Good morning.

This week we’re going to give the AI skeptics and haters a voice. And believe me, their numbers are growing every week. Entire YouTube channels are dedicated to “debunking AI”.

Part of this week’s financial turmoil was also based on some doubts being raised about AI

Throwing Bags of Money at AI

Companies like OpenAI and Anthropic are collecting billions of dollars from investors on the wave of the AI hype. NVIDIA is selling their GPUs (graphical cards) like hot cakes. So much in fact they have become the most valuable company in the world.

All this money in search of a promise that hasn’t been delivered yet and according to the critics is unattainable : artificial general intelligence (AGI) and - dare to say - SUPERintelligence or ASI.

If you’re still not clear : AGI, or Artificial General Intelligence, is like a super-smart computer that can think and learn just like humans, able to handle any task we can. ASI, or Artificial Superintelligence, goes even further - it's an AI system that's way smarter than any human, capable of solving problems and coming up with ideas beyond what we can even imagine.

Pioneers in the field, people like Ilya Sutskever and Andrej Karpathy (both founding fathers of OpenAI) they believe superintelligence is a matter of processing power and time. They’re debating AI safety on podcasts because superintelligence is definitely arriving and we should become ready. Fast. Because humanity is in danger. In fact, Sutskever has left OpenAI in a conflict on safety and has now founded its own company which mission is to create a secure superintelligence.

But - say the skeptics - these dudes are talking about AI safety to extract more money from investors. It’s all about money.

Main Criticisms

So why is there this counter-move against AI , while we all can see what these large language models can do for us.

The main criticism is that computers are computers and at their core a large language model is a pattern recognition system. Dumb as a rock but because they’re good at predicting the next character they SEEM smart.

These models cannot achieve true intelligence because they don’t understand what they’re feeding back to our questions.

They’re also not grounded in the real world - lacking experience of the real world which is a fundamental barrier to developing genuine intelligence.

Also, once they’re trained - they cannot update their knowledge in real-time. Humans can adapt based on new information. LLMs are not doing that.

They also lack “intellectual breadth” - emotional intelligence or spatial reasoning. Those also are a part of being considered “intelligent”.

On top of that - they seem to be inconsistent and prone to hallucinations.

The Main Counter-Argument : Emergence

The main counter-argument against the critics is that these models show emergent properties.

As these AI models get bigger and more complex, they start showing abilities that weren't specifically programmed. It's as if by learning to handle language really well, they accidentally learn other skills too - like solving math problems or understanding jokes. This happens because the model picks up on subtle patterns and connections in the data it was trained on, leading to new and sometimes surprising capabilities.

AI researchers at OpenAI are still discovering things GPT4 can do that are unexpected.

Consciousness

And that’s the whole idea of AI’s becoming “conscious”. As they grow in size and become more complex - it’s possible that enough virtual neurons are connected that the machine becomes aware of itself.

But we are now in uncharted territory because we don’t know what consciousness IS.

Anyway, there’s something to think about this weekend. What side are you on ?

Welcome to the Blacklynx Brief

AI News
Is Google Search In Trouble ?

• Google DeepMind's experimental Gemini 1.5 Pro has claimed the top spot on the AI Chatbot Arena leaderboard with a score of 1300, surpassing OpenAI's GPT-4o and Anthropic's Claude-3.5. This unexpected rise suggests Google may have taken the lead in the LLM space, pending further updates.

• Black Forest Labs launched FLUX.1, a suite of AI image generation models in three variants: [pro], [dev], and [schnell], rivaling leaders like Midjourney and DALL-E 3. The models are open-source, with the [dev] variant available for non-commercial use and [schnell] optimized for personal use. They also teased an upcoming text-to-video generation model.

• Stability AI introduced Stable Fast 3D, an AI model that generates high-quality 3D assets from a single image in just 0.5 seconds. Available on Hugging Face and through their API, this model drastically reduces generation time from 10 minutes to half a second, potentially transforming industries like gaming, e-commerce, and architecture.

• Meta is reportedly offering millions to celebrities like Awkwafina, Judi Dench, and Keegan-Michael Key to use their voices in upcoming AI projects across its platforms, including Facebook, Instagram, and Meta Ray-Ban smart glasses. Contracts are temporary with options to renew, aiming to enhance user engagement and growth.

• Google signed a non-exclusive licensing agreement with AI startup Character AI for its language model technology and reabsorbed the startup's co-founders and key talent. Character AI will shift to open-source models like Meta's Llama 3.1, while about 30 employees join Google's Gemini AI efforts, reflecting the intense competition for top AI talent.

• Tesla filed a patent for an AI-powered vision system that uses camera inputs to create detailed 3D environment information without LiDAR or radar. Designed for autonomous robots and vehicles, this system could enable real-time decision-making and navigation, potentially transforming robotics with more efficient and adaptable solutions.

• Three key leaders at OpenAI, including co-founder John Schulman, co-founder Greg Brockman, and product leader Peter Deng, are departing or taking leave. Schulman has joined rival Anthropic, Brockman is on extended leave, and Deng has left the company. This follows other high-profile exits and occurs amid increased competition from Anthropic and Meta AI.

• Elon Musk has filed a lawsuit against OpenAI and CEO Sam Altman, alleging manipulation and deceit in the company's transition from non-profit to for-profit status, including violations of federal racketeering laws. OpenAI denies the allegations. The lawsuit follows a previous case Musk dropped and comes amid leadership turmoil at OpenAI.

• Leaked documents reveal that Nvidia has been scraping millions of videos daily from platforms like YouTube and Netflix to train its AI model, Project Cosmos. Despite internal concerns about copyright and ethics, Nvidia claims its practices comply with copyright law. This project could significantly impact various AI applications, but the approach raises ethical and legal questions.

• OpenAI-backed Figure AI unveiled Figure 02, an advanced humanoid robot designed for autonomous work in complex environments like factories. The robot uses OpenAI's models for speech reasoning, has 360-degree vision from six RGB cameras, and boasts a 20-hour runtime with a custom battery pack. This positions Figure 02 as a strong competitor to Tesla's Optimus.

• ByteDance launched Jimeng AI, a text-to-video app for Chinese users, competing with OpenAI's yet-to-be-released Sora AI video model. Available on both iOS and Android, Jimeng AI allows users to create AI-generated videos and images through a subscription model. This move intensifies the AI video generation race in China, leveraging ByteDance's extensive resources and data.

• Researchers developed a new Listening-While-Speaking Language Model (LSLM) that enables AI to listen and speak simultaneously, allowing for real-time, interactive speech-based conversations. The model uses advanced techniques for speech generation and audio input, enhancing the naturalness and responsiveness of AI interactions. This advancement could significantly improve human-AI communication.

• A mysterious AI model, dubbed "anonymous-chatbot," has surfaced in the LMSYS Chatbot Arena, sparking rumors it might be OpenAI's next major AI, codenamed 'Strawberry'. This model shows advanced reasoning capabilities beyond GPT-4o, and speculation is fueled by a cryptic tweet from Sam Altman.

• Researchers developed an AI model that can predict major diseases like heart conditions, diabetes, and cancer with 95% accuracy, outperforming existing methods. Using patient data and a smart algorithm, it aids early diagnosis and personalized healthcare, potentially revolutionizing treatment planning and saving millions of lives.

• OpenAI is leading a $60 million Series B funding round for Opal, known for high-end webcams, to develop AI-powered consumer devices. This partnership aims to integrate OpenAI’s voice AI models into new devices, indicating OpenAI's strategic move to bring advanced AI technology directly to consumers.

Quickfire News

• Google introduced new AI features for Chrome: Google Lens for desktop, Tab compare for product comparisons, and improved browsing history search.

• GitHub launched GitHub Models: platform for accessing and experimenting with AI models, includes a playground, Codespaces integration, and deployment.

• Healx raised $47 million in Series C funding: regulatory clearance for Phase 2 clinical trials for a new rare disease treatment in the U.S.

• Google faced backlash over its Gemini AI Olympics-themed ad: critics argue it promotes overreliance on AI tools at the expense of children's learning and creativity.

• Microsoft listed OpenAI as a competitor in its annual report: despite their partnership and Microsoft's significant investment.

• Character AI open-sourced Prompt Poet: aims to revolutionize AI interactions in production environments.

• Adobe Illustrator introduced AI-powered generative vector features: allows creation of complex vector art and patterns from text prompts.

• UC Berkeley unveiled the 'Berkeley Humanoid': a 2 ft. 9 in. humanoid research platform for learning-based control and robust locomotion.

• Britain canceled $1.7 billion worth of computing infrastructure projects: includes plans for an exascale computer, impacting AI ambitions.

• China's Cixin Technology unveiled the Cixin P1: a 6nm Arm-based 12-core AI chip for PCs, capable of 45 TOPS and compatible with multiple operating systems.

• Microsoft modified its AI partnership with UAE-based G42: opted to lease AI products instead of transferring sensitive technology due to concerns over G42's ties to China.

• An autonomous robot dentist performed the world's first human dental procedure: prepared a tooth for a crown in about 15 minutes, faster than human dentists.

• Groq raised $640 million in Series D funding: aims to scale AI inference capacity and accelerate development of next-gen LPU technology.

• Meta announced the Llama 3.1 Impact Grants program: offers up to $2 million for projects using Llama 3.1 to address social challenges.

• Caristo Diagnostics developed new AI technology: detects hidden heart attack risk by analyzing CT scans for coronary inflammation.

• OpenAI reportedly built a text watermarking tool: detects ChatGPT-written content with 99.9% effectiveness but not released publicly.

• Tencent joined a $300+ million funding round for Chinese AI startup Moonshot: valuing the company at $3.3 billion amid rising competition in China's AI sector.

• Secretaries of state from five U.S. states urged Elon Musk to address misinformation: spread by X's AI chatbot Grok regarding the upcoming November election.

• OpenAI introduced Structured Outputs for its API: allows developers to ensure AI-generated outputs match specific JSON schemas.

• Colorado released an AI roadmap for schools: provides guidelines on integrating AI into education policy and curricula.

• ProRata AI raised $25 million: partners with major media companies to develop an AI chatbot with a revenue-sharing model based on content attribution.

• Cleveland Clinic launched a Quantum-AI Biomedical Frontiers Fellowship Program: integrates quantum computing and AI into healthcare research.

• Japanese scientists developed a simplified EUV lithography tool: aims for cheaper chip production, potentially revolutionizing AI hardware manufacturing.

• Sonova introduced Sphere Infinio: the first hearing aid utilizing real-time AI to improve speech clarity from background noise.

• Reddit announced plans to test AI-generated summaries at the top of search result pages: uses a combination of first-party and third-party technology to enhance content discovery.

• Mistral AI launched model customization: includes an early version of Agents and a stable SDK to simplify generative AI application development.

• Google announced a new Gemini AI-powered TV streamer: replaces Chromecast with advanced smart home and entertainment features.

• Audible began testing AI-powered search feature "Maven": provides personalized audiobook recommendations based on user queries.

• Wendy's introduced Spanish AI ordering in Florida drive-thrus: enhances accessibility for Spanish-speaking customers.

• Intel reportedly declined an opportunity to invest in OpenAI in 2017: missed early entry into the AI market due to doubts about AI's near-term potential.

• Verizon deployed AI and machine learning: predicts and prevents fiber cuts, enhancing network protection efforts.

Closing Thoughts

That’s it for us this week.

If you find any value from this newsletter : make sure to send the link below to your colleagues and friends :

https://www.blacklynxbrief.com/subscribe