AI's Trillion Dollar Time Bomb

Plus, an AI cheat sheet, cool new AI models, and a cigarette-eating robot

Author

Steve Brown
July 25, 2024

In partnership with

Subscribe to Synthetic

The most interesting and relevant AI news and analysis of the week

This Week in AI

Much of the buzz in the robotics world has been about human-like robots from Figure, Tesla, and Agility Robotics. Along comes a new robot called VERO with vacuum cleaner feet and an appetite for cigarette butts. Smokers discard an estimated 4.5 TRILLION cigarette butts every single year (come on, smokers, you know better). This video shows VERO in action. And if cute robots are your jam, check out the new robots coming out of Disney Research. đŸ€–

Recently, Synthetic has warned that if the education sector is slow to adopt AI technology and fundamentally transform the teaching and learning process, someone else will do it for them. Enter Eureka Labs, Andrej Karpathy’s new startup with a mission to build an “AI native” school. Educators shouldn’t panic yet. Eureka’s vision is for AI to support teachers, not replace them. Eureka’s first focus is an undergraduate training course to teach people how to build
.<drumroll please>
.artificial intelligence.

The Atlantic’s excellent technology writer, Charlie Warzel, describes Silicon Valley’s AI ‘audacity crisis’ in which tech executives, high on manifest destiny and the smell of their own farts, push aside concerns over climate-wrecking energy consumption and widespread job destruction and march on in the name of progress. Synthetic is generally an optimistic rag, so it was time for a bit of balance to be restored to the force. 😀

Quick Hits

  • No AI For You! - Citing an unpredictable regulatory environment, Meta cancels plans to release gen AI models in the EU and Brazil.

  • VideoPoet - Google Research debuts a new video and audio generation model.

  • Reddit Gets Nasty - Reddit now blocks all AI bots. Unless they pay.

  • AI Boyfriend Business is Booming - Women are turning to specialist chatbots for companionship and someone who will listen.

  • Apple Intelligence Here Now - Some AI features are available in the first public beta of iOS 18 and iPhone users can try it now.đŸ“±

Video: AI’s Trillion Dollar Time Bomb

Sequoia Capital was the first to identify what it termed “AI’s $600 billion question,” the gap between the amount invested in AI and the return on that investment so far. A Goldman Sachs report asks if generative AI has had “too much spend, too little benefit” and estimates the gap at $1 trillion.

This CNBC video explores the issue and asks if AI hyperscalers are overinvesting and experiencing “irrational exuberance.” Not that you asked, but Synthetic’s opinion is that frontier AI requires significant forward-looking capital investment, and the true benefits, impact, and returns will not be felt until the 2026-2027 timeframe. The AI race is not a game for the faint of heart or the short of cash.

AI Tech and Innovation

Llama 3.1 405B is here - the most capable open-source model yet

With another entry for the “worst name for an AI model” contest, this week Meta introduced Llama 3.1 405B, their latest foundational multimodal model. 405B, named for its 405 billion parameters, joins updated versions of 70B and 8B to form the new Llama family. Benchmarks presented by Meta show 405B performing well against GPT-4, GPT4-o, and Claude Sonnet, closed-source models from OpenAI and Anthropic. You can download and try the model here. It’s also available now on AWS. 🩙

Presenting at ICML 2024, a Mecca for AI researchers, Google DeepMind is publishing over 80 new AI research papers this week. Among the wide-ranging topics covered is an update of DeepMind’s thinking on a framework to categorize the capabilities and behaviors of artificial general intelligence (AGI) models and their precursors. In the paper, Google DeepMind defines five capability levels of AGI (Emerging, Competent, Expert, Virtuoso, Superhuman) and goes on to define five levels of autonomy for AI agents:

  1. Tool - Human in control; AI automates mundane sub-tasks

  2. Consultant - Human invokes the AI; AI plays a substantive role

  3. Collaborator - Co-equal human/AI involvement

  4. Expert - AI drives the interaction; Human provides guidance or performs sub-tasks

  5. Agent - Fully autonomous AI

You can download the full paper here. It’s well worth reading (and some of Synthetic’s friends wrote it, so that’s extra cool).

OpenAI’s new model, GPT-4o Mini (their marketing department really needs to up their game), is available now and costs significantly less to use than full-sized models. It competes with Anthropic’s Claude 3 Haiku and Google’s Gemini Flash models, which are designed to be fast and light, provide quick responses, and run on a small computing footprint.

The New York Times reports that research from the Data Provenance Initiative indicates a drop in the content made available to the collections used to train artificial intelligence. Based on LLMs, frontier models are expected to benefit from both compute and data scaling, at least for the next two to three generations. AI research labs will have to pony up for access to the high-quality data needed to train next-generation models.

Reuters reported rumors that OpenAI is not only hiring former Google employees to develop an in-house AI server chip but is also courting chipmaker Broadcom in an attempt to lessen the market’s reliance on Nvidia for leading-edge AI hardware and expand access to AI benefits.

AI Insights, Analysis, and Opinion

❝

“The Industrial Revolution came for farmer’s jobs (and some factory jobs). The AI Revolution is coming for desk workers.”

Rex Woodbury, writing in Digital Native (see article below)

Digital Native’s Rex Woodbury explores how large language models will transform knowledge work in the coming years. Citing research from the U.S. Bureau of Labor Statistics, Goldman Sachs, Bain, and Barron’s, Woodbury picks the service economy sectors most easily transformed by language models - sectors that can be disrupted with today’s technology, not the AI models we hope will come in the future. Finally, he shares a simple three-step formula for startups to create value and build a successful business focused on the services sector.

This is a great AI primer for Synthetic readers who are still figuring out the difference between Generative AI, Artificial General Intelligence, and Retrieval Augmented Generation, a.k.a. RAG. The article defines key AI terms such as parameters, tokens, and transformers and concludes with some explanation of key AI hardware building blocks.

According to a new Capgemini report on “Harnessing the Value of Generative AI,” a massive 82% of companies plan to deploy and integrate AI agents into their workflows in the next 1-3 years. Agentic AI is all the rage at the moment, and AI researchers are racing to launch the first useful autonomous AI agents. Capgemini Chief Innovation Officer Pascal Brier expects agents to hit the market in 2025 and quickly collaborate across company departments. For example, a marketing agent might autonomously run a new ad campaign concept by a legal agent for review.

Fund manager James Anderson, an early investor in Tesla and Amazon, says Nvidia could be worth $49 trillion a decade from now based on “persistent exponential progress, the competitive advantage in hardware and software, and the culture and leadership.“ Anderson gives this outcome a 10-15% probability. A hedge fund manager recently predicted Nvidia stock could surge to $250 by the end of 2024, yielding a $6 trillion market cap. With money rotating out of the tech sector, that seems unlikely.

Toolkit for the Future

Your Brilliant Business Idea Just Got a New Best Friend

Got a business idea? Any idea? We're not picky. Big, small, "I thought of this in the shower" type stuff–we want it all. Whether you're dreaming of building an empire or just figuring out how to stop shuffling spreadsheets, we're here for it.

Our AI Ideas Generator asks you 3 questions and emails you a custom-built report of AI-powered solutions unique to your business.

Imagine having a hyper-intelligent, never-sleeps, doesn't-need-coffee AI solutions machine at your beck and call. That's our AI Ideas Generator. It takes your business conundrum, shakes it up with some LLM magic and–voila!--emails you a bespoke report of AI-powered solutions.

Outsmart, Outpace, Outdo: Whether you're aiming to leapfrog the competition or just be best-in-class in your industry, our custom AI solutions have you covered.

AI extracts information from your receipts and organizes it in an easy-to-search database alongside IRS-approved, secure image scans. Capture receipts on the gođŸ“±, forward receipts 📧, use the web portal, or mail in a pile of physical receiptsđŸ§Ÿ. Try it free for 30 days.

Extract data from websites and place it in a spreadsheet that fills itself and notifies you of any changes. Pre-built robots make it a breeze to get started. Extract whatever you need: job listings, property details, TikTok videos, hotel info, or competitor pricing. Free to try. No coding required.

A new set of generative AI-powered tools to transform images, remove backgrounds and watermarks, create backgrounds, and visualize products. Deliver one-of-a-kind visual experiences and better engagement on the web. Free to try.

Keep your sales team focused on customers during meetings. Laxis captures attendee comments verbatim and flags items for follow up. It auto-generates meeting summaries and follow-up emails in seconds, quickly identifying customer requirements, pain points, and action items. Free to try.

This versatile AI assistant helps content creators to write text, edit content, change writing tone, create images, generate audio, easily translate text into other languages, quickly summarize hours-long YouTube videos, generate memes, and more. Free to try.