About This Week’s Covers
This week celebrates Halloween so I used the Ideogram API to generate 33 category covers in the style of Francis Bacon, one of the most haunting,terrifying, and macabre of my favorite painters. For the main cover image, the result from Ideogram was so disturbing that I don’t feel comfortable using it! I’d say that makes for a successful Halloween prompt test. The prompt I used was: “A disturbing painting symbolizing the death of the web browser in the style of Francis Bacon.” However, Ideogram interpreted “browser” as a person, leading to a result that’s truly Halloween-worthy—I included it below the fold (see down a bit) for a legit Halloween spook.
As usual, I did a bake-off with MidJourney and Flux to see how they compare, and in this case, MidJourney’s weakness was a strength. MJ’s idea from my prompt was a basic yet spooky skeleton, so that became the cover this week.
The API-driven covers were all over the place, and general not for the faint of heart, especially in a corporate setting. You can find a few examples below the macabre Ideogram cover image. Happy Halloween. AI is indeed the stuff of nightmares! Next week, I’ll do something lighter, LOL.
Ideogram went a bit too overboard with my Halloween request (at least for a general audience):
Examples of Ideogram’s automated API driven covers for categories in the style of Francis Bacon. Definitely Halloween-worthy and pretty good at combining the category theme with Bacon’s style. It’s mostly a question of appropriateness for general audiences, and I’d rather that be left to the user than the platform denying the request.
This Week’s Executive Summaries
Since September 2023 (in a presentation to the National Association of Broadcasters – slide 14), I’ve been yelling constantly that we need to prepare for the death of the web page view. It always makes me feel vindicated when I see articles echoing this idea. But I’m not sure what’s going to happen. See also, an article I wrote in November 2023. “The AI Future: Exploring the Adjacent Possible with Emerging AI Solutions”
MIT Technology Review: AI Search Risks Breaking the Web, Demands Proactive Industry Action
The rise of AI search engines like Perplexity AI and Google Gemini, which summarize and source information from across the web, threatens to disrupt the digital content economy. These tools streamline user access to knowledge but reduce traffic to the original sources that depend on visits for revenue. A lawsuit by News Corp against Perplexity AI underscores mounting tensions, as publishers argue AI systems profit unfairly from their content. While current copyright laws offer limited solutions, governments may impose blunt regulations like mandatory bargaining, which risks favoring legacy media and undermining the open web’s principles. AI firms face a narrow window to address these issues by creating equitable systems that reward all creators for contributing valuable content. Such frameworks could prevent the dominance of a few major players while fostering a sustainable digital ecosystem.
Technologyreview
ChatGPT Integrates Web Search for Real-Time Information
OpenAI has upgraded ChatGPT with internet search functionality, enabling users to access up-to-date answers directly within the chat interface. Aka the death of the web browser part I. Users can ask conversational questions and receive accurate, timely responses from reliable websites. GPT search also integrates with direct data partners for sports scores, news updates, stock prices, and weather forecasts. Available immediately for Plus and Team users, with a broader rollout planned, the tool partners with major news providers like Vox Media, Le Monde, and Reuters. Responses include clickable source links. The search model builds on GPT-4o, with plans for further expansion into shopping, travel, and voice integration.
openai
Google Revenue Up, Cost per AI Query Down
Alphabet Revenue Jumps 15% to $88.3 Billion – The New York Times
https://www.nytimes.com/2024/10/29/technology/alphabet-google-earnings.html
“Wild stat: Google’s cost per query for AI search results dropped by 90% — while doubling the model size! AI answers now reaches 1 billion users a month — so those savings compound quickly at Google scale.
https://twitter.com/bilawalsidhu/status/1851391059781153105
Elon’s X Now Has the World’s Largest AI Supercomputer
NVIDIA Ethernet Networking Accelerates World’s Largest AI Supercomputer, Built by xAI | NVIDIA Newsroom
https://nvidianews.nvidia.com/news/spectrum-x-ethernet-networking-xai-colossus
“NVIDIA’S JENSEN HUANG: ELON IS SUPERHUMAN His comments came after xAI set a jaw-dropping record with its “Colossus” supercomputer. In just 19 days, xAI installed 100,000 Nvidia H100 chips to power Grok. Huang, rarely one to gush, said the installation speed was
https://twitter.com/MarioNawfal/status/1850560225545277893
While Tesla Promises Self-Driving Taxis, Waymo is Servicing 150,000 Full Autonomous Taxi Rides Per Week
“Waymo One is now providing over 150,000 paid trips and driving over 1 million fully autonomous miles every week. The future is here, it’s growing, and it’s taking riders safely to their destinations every day.
https://twitter.com/Waymo/status/1851365483972538407
Report on From CitiGroup – AGI By 2029 (WHOA)
“Slightly schizophrenic report from Citigroup. AGI is arriving in 2029, with ASI soon after. But don’t worry: work on your critical thinking, problem solving, communication and literacy skills for a durable competitive advantage.
https://twitter.com/ben_j_todd/status/1848691312704229681
Related: Silicon Valley Takes AGI Seriously—Washington Should Too | TIME
https://time.com/7093792/ai-artificial-general-intelligence-risks
Google Is Using AI to Crush Productivity Goals
Google now uses AI to write 25% of its new code — Alphabet CEO Sundar Pichai underlines the company’s role in the AI industry amidst strong Q3 24 financials | Tom’s Hardware
https://www.tomshardware.com/tech-industry/artificial-intelligence/google-now-uses-ai-to-write-25-percent-of-its-new-code-alphabet-ceo-sundar-pichai-underlines-the-companys-role-in-the-ai-industry-amidst-strong-q3-24-financials
Best AI Visuals and Charts: Week Ending 11/01/2024
“I don’t know if we live in a Matrix, but I know for sure that robots will spend most of their lives in simulation. Let machines train machines. I’m excited to introduce DexMimicGen, a massive-scale synthetic data generator that enables a humanoid robot to learn complex skills
https://twitter.com/DrJimFan/status/1852383627738239324
“Everything you do wearing AR glasses will be reconstructed in world space. Pretty impressive to see body and hand motion recovered with such detail! And the visualizations with 3D Gaussian Splat maps are a nice touch to showcase the world space accuracy:
https://twitter.com/bilawalsidhu/status/1850224887035789748
EchoMimic – a Hugging Face Space by fffiloni
Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning
https://huggingface.co/spaces/fffiloni/EchoMimic
“Easily the most impressive uncut autonomous video I’ve ever seen, from @physical_int
https://twitter.com/chris_j_paxton/status/1852047463978254460
“Playing billiards with the Snapchat AR spectacles on. There’s a lens that overlays where I need to hit the ball to make the shot. Is this cheating or nah? 🤓 (Too bad I scratched anyway)
https://twitter.com/MKBHD/status/1851356704145096935
“We’re excited to see creators like Ken @ODonnellKen push the boundaries of what’s possible using our markerless motion capture technology! Whether you’re an indie studio or a creative team, Move AI empowers you to bring your ideas to life at scale. 🌍✨ #MoveAI
https://twitter.com/MoveAI_/status/1848696832378966359
Top 67 Links of The Week – Organized by Category
Agents and Copilots
“Significant development for the power of AI Agents, they are learning to use GUIs. Agent-S uses a computer like a human to solve diverse desktop tasks on different systems, by breaking them down into bite-sized actions and remembering what works. Experience-augmented
AI researchers put LLMs into a Minecraft server and said Claude Opus was a harmless goofball, but Sonnet was terrifying – “the closest thing I’ve seen to Bostrom-style catastrophic AI misalignment ‘irl’.” : ClaudeAI
Bringing developer choice to Copilot with Anthropic’s Claude 3.5 Sonnet, Google’s Gemini 1.5 Pro, and OpenAI’s o1-preview – The GitHub Blog
“Gemini 1.5 Pro in Github Copilot as well. 👏👏 In addition to Claude 3.5 Sonnet and OpenAI’s o1-preview.” / X
Google is reportedly developing a ‘computer-using agent’ AI system – The Verge
Bringing developer choice to Copilot with Anthropic’s Claude 3.5 Sonnet, Google’s Gemini 1.5 Pro, and OpenAI’s o1-preview – The GitHub Blog
GitHub Copilot will support models from Anthropic, Google, and OpenAI – The Verge
AGI (Artificial General Intelligence)
“Just started at @OpenAI and I couldn’t be more excited to join at this pivotal moment of safe AGI development! Met so many old friends already, talent density of this place is just insane!! Thank you all for the warm welcome, and in particular @sama. Now let the unicorns fly!
Audio
“We’re excited to introduce Personas 🎤 Personas let you save the essence of a song – vocals, style, vibe – and reimagine it across your creations. Want to bring your favorite vocals or style into a new song? Here’s how: 1⃣ Find a song you love, click Create then Make a Persona
OpenAI’s Whisper transcription tool has hallucination issues, researchers say | TechCrunch
Augmented and Virtual Reality (AR/VR)
PF3plat : Pose-Free Feed-Forward 3D Gaussian Splatting
LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias
NoPoSplat
Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images
Autonomous Vehicles
Elon Musk finally admits Tesla’s HW3 might not support full self-driving | Electrek
Business and Enterprise
LinkedIn launches its first AI agent to take on the role of job recruiters | TechCrunch
“I keep hearing from executives that they expect that a new generation of “AI natives” will show them how to use AI. I think this is a mistake: 1) Our research shows younger people do not really get AI or how to integrate into work 2) Experienced managers are often good prompters
“New longitudinal GenAI adoption survey by my colleagues at Wharton (not me), surveying 800 senior managers at big firms, finds usage doubled in a year: 72% use at least once a week. The vast majority report positive impacts. The pace of real world AI adoption continues rapidly.
Chips, Hardware, and Infrastructure
Why a Memphis Community Is Fighting Elon Musk’s Supercomputer – The New York Times
Nvidia overtakes Apple as world’s most valuable company | Reuters
Ethics/Legal/Security
The Pentagon Wants to Use AI to Create Deepfake Internet Users
In my experience, people who think this concern is minor or not commensurate with (even low likelihood of) more straightforward AI doom… just don’t mind the idea of a one world nanny state government.
But I believe it can get really bad, yes, commensurate with extinction.
Google now uses AI to write 25% of its new code — Alphabet CEO Sundar Pichai underlines the company’s role in the AI industry amidst strong Q3 24 financials | Tom’s Hardware
Google just gave its AI access to Search, hours before OpenAI launched ChatGPT Search | VentureBeat
Gemini
Google plans next Gemini 2.0 model for December release – The Verge
Imagery
A mysterious new image generation model has appeared | TechCrunch
“👀 Hold up: A mysterious ‘red_panda’ model just topped the text-to-image leaderboard, outperforming Black Forest Labs & Midjourney with 1172 ELO & 72% win rate ! Who’s behind this challenger? 🐼
How X users earn thousands from US election misinformation and AI images
Ideogram
“Ideogram launched Canvas, a creative platform for AI image generation and editing. The system features Magic Fill for precise regional editing and Extend for expanding images beyond borders Both tools maintain a consistent style across modifications
Meta
For Llama updates please see the Open Source category
Meta Develops AI Search Engine to Lessen Reliance on Google, Microsoft — The Information
“Meta spent as much as the Manhattan Project on GPUs in inflation-adjusted dollars. Really curious how long they will continue to release open models with that sort of capital expenditure and models that are getting more economically valuable.
Multimodality
“AI has achieved ‘scent teleportation’ 🤯 Photos, videos, and audio have been digital for decades, and now smell is joining the digitization. Here’s how Osmo, a 2-year-old American startup, made it happen: > AI analyzed, digitized, and reproduced the smell of plum, creating an
“Well, we actually did it. We digitized scent. A fresh summer plum was the first fruit and scent to be fully digitized and reprinted with no human intervention. It smells great. Holy moly, I’m still processing the magnitude of what we’ve done. And yet, it feels like as we cross” / X
OpenAI
“🌐 Introducing ChatGPT search 🌐 ChatGPT can now search the web in a much better way than before so you get fast, timely answers with links to relevant web sources.
“chatgpt search vs perplexity
OpenAI’s search engine is now live in ChatGPT – The Verge
OpenAI CFO Says 75% of Its Revenue Comes From Paying Consumers
ChatGPT Topped 3 Billion Visits in September | Similarweb
“🌐 Introducing ChatGPT search 🌐 ChatGPT can now search the web in a much better way than before so you get fast, timely answers with links to relevant web sources.
Open Source
“Great to visit one of our data centers where we’re training Llama 4 models on a cluster bigger than 100K H100’s! So proud of the incredible work we’re doing to advance our products, the AI field and the open source community. We’re hiring top researchers to work on reasoning,
“Training Llama 4 on 100k H100s.” / X
Perplexity
“We released Spaces last week, and have already been shipping updates based on your feedback. Here’s what’s improved: 1. Free users can now upload 5 files per Space 2. Enhanced custom instructions: Spaces AI assistant now applies them consistently across all threads 3. More
“Perplexity Sports! Starting with NFL widgets (built in house) where you can get in depth game summaries, stats and comparisons across players and teams! NBA, EPL, La Liga, champions league, tennis, cricket, etc will all be coming soon!
“Introducing Perplexity Supply. Quality goods, thoughtfully designed for curious minds.
Publishing
“Voice Actors have collectively earned over $1M on ElevenLabs since we added payouts in February. Thank you to everyone who contributed their voice so creators have limitless possibilities. Hear some stories behind the voices here:
“AI search engines threaten to break the web’s ecosystem by bypassing source websites. Interesting opinion piece in the @techreview: The AI industry must build fair reward systems before governments step in – by @opensauceAI
Meta is reportedly developing a search engine for its chatbot
“chatgpt search vs perplexity
OpenAI’s search engine is now live in ChatGPT – The Verge
OpenAI CFO Says 75% of Its Revenue Comes From Paying Consumers
ChatGPT Topped 3 Billion Visits in September | Similarweb
“🌐 Introducing ChatGPT search 🌐 ChatGPT can now search the web in a much better way than before so you get fast, timely answers with links to relevant web sources.
Introducing ChatGPT search | OpenAI
“🌐 Introducing ChatGPT search 🌐 ChatGPT can now search the web in a much better way than before so you get fast, timely answers with links to relevant web sources.
(RAG) Retrieval-Augmented Generation
“2mn context window LLMs are already facts of today. 100mn context window are in very advanced stage of work in progress. There are some popular papers from big labs that have found ways to do 1-Bn context window.. So RAG’s long-term (5Years+) future is uncertain.
Robotics and Embodiment
“Clone Robotics showed off Torso, a bimanual android powered by artificial muscles. The humanoid system features anatomically accurate joints across shoulders, neck, and arms, with a valve control system integrated into its ribcage structure.
Advancing embodied AI through progress in touch perception, dexterity, and human-robot interaction
Meta is making a robot hand that can ‘feel’ touch | TechCrunch
“@torcrobotics completes driverless #trucking tests, readies for 2027 commercialization
Atlas Goes Hands On – YouTube
Elon Musk: 10 billion humanoid robots by 2040 at $20K-$25K each
Video News
“AI startup Genmo just launched Mochi 1, a new open-source video generation model The startup claims to rival closed competitors like Runway, Pika, and Kling — while being freely available to developers and researchers Another win for open-source
“Genmo Mochi 1 outperforms Runway, Kling, Luma and Pika’s video generation models according to community votes. And it’s 100% open source, Apache 2.0 licensed. Next stop: #1!” / X
Runway
“Or quickly zoom out to reveal new context and story. (7/8)
“Advanced Camera Control is now available for Gen-3 Alpha Turbo. Choose both the direction and intensity of how you move through your scenes for even more intention in every shot. (1/8)
Twitter/X/Grok
“Grok 2’s multimodal image input is quite good. I am impressed that both it and GPT-4o caught the “vibe” that this was taken in Asia (GPT-4o even guessed Japan). It was taken in Osaka, all location data was removed from the image, and there is no visible signage.
Elon Musk’s xAI adds image understanding capabilities to Grok | TechCrunch
Inside the World’s Largest AI Supercluster xAI Colossus – YouTubehttps://www.youtube.com/watch?v=Jf8EPSBZU7Y
Leave a Reply