• Synthetic
  • Posts
  • OpenAI's Five Levels of AI and Secret Reasoning Technology

OpenAI's Five Levels of AI and Secret Reasoning Technology

Plus, AI Accurately Predicts Alzheimer's Progression

In partnership with

Subscribe to Synthetic

This Week in AI

Last week, OpenAI unveiled a new classification framework for AI systems at an all-hands meeting with employees. A spokesperson later shared the framework with a Bloomberg reporter. In OpenAI’s model, we are almost at step two on a five-step journey towards AGI, with GPT-5 (or whatever it is called when released) likely delivering at level 2 and paving the way for a rapid progression to level 3 (reasoning is a key capability of agentic AI).

Levels
1: Chatbots - AI with natural conversation abilities
2: Reasoners - Human-level problem-solving across a broad range of topics
3: Agents - Systems that can take actions independently or from human instruction
4: Innovators - AI that can aid in the invention of new ideas
5: Organizations - AI that can do the work of an organization

This framework feels somewhat similar to the five levels of autonomous driving and contrasts with the excellent framework proposed by DeepMind last year that sets performance and generality goals for both narrow and general AIs.

The government and the Justice Department are watching while tech giants jostle for position in the AI race. With AI talent at a premium, the practice of ‘acqui-hiring’ has become widespread. Microsoft gobbled up Inflection, and more recently, Amazon has absorbed most of the talent at Adept. Several U.S. senators are paying attention, and the DOJ and FTC are watching closely. Expect investigations to be announced soon, focused on anti-competitive behavior in the AI marketplace. The UK already announced a probe into Microsoft’s acqui-hire of Inflection AI.

From this week’s department of furious backpedaling comes the tale of a software company that caused an uproar with a poorly considered HR policy and press release. Lattice Software declared that AI workers would have managers and be held accountable to specific goals and standards to assess their performance like every other employee. The pushback from employees was strong: “Treating AI agents as employees disrespects the humanity of your real employees.”

Quick hits

  • Sneaky Gemini - Google Gemini was caught scanning Google Drive user files without permission.

  • Robot Brain Unicorn - Startup Skild AI emerged from stealth with $300m in funding from Jeff Bezos and Softbank, a $1.5 billion valuation, and a focus on building AGI and body-agnostic brains for robots.

  • World Labs Already Worth $1 Billion - Startup founded by Fei-Fei Li, the ‘godmother of AI,’ will use human-like visual data processing to understand the physical world and build advanced reasoning AI. The company is just four months old.

  • Battle for AI Search Supremacy - Can AI-powered “answer engines” like Perplexity replace Google’s 10 blue links?

  • People Think ChatGPT is Conscious - Survey says!! Two-thirds of people believe (erroneously) that it has feelings and memories.

  • AI Finally Shakes Up Education - A detailed report by Bond predicts AI will significantly change universities.

Video: The Moment We Stopped Understanding AI

This excellent explainer from Welch Labs explores how most of today’s AIs work. It also reviews some of the history of AI and how today’s advanced large language models are built on technology developed in the 1950s, ‘60s, ‘70s, and ‘80s. AlexNET signaled a major breakthrough in the field of deep learning, the most popular approach for AI today. This video explains why and asks what may come next. Viewing tip: If you find some concepts too advanced for you, sit tight, and you’ll learn more as the video progresses. (17 mins)

AI Tech and Innovation

Reuters broke the news that OpenAI continues to work on improved reasoning technology, which is seen as a critical breakthrough needed to build more intelligent AI systems. Last year, amid the is-he-isn’t-he-CEO debacle, rumors swirled of a reasoning tech codenamed Q-STaR. (STaR = Self-Taught Reasoner). Reasoners are important for two reasons:

1) To improve an AI’s abilities to solve problems and answer challenging questions accurately

2) To enable AIs to self-improve by conducting deep research and generating new training data sets for next-generation AIs.

Details of ‘Strawberry’ remain a tightly held secret, even within the company.

Legendary chip designer Jim Keller (formerly of Intel, AMD, and Tesla), wants to reduce the cost of AI applications with new chips that cost less and consume less energy than those made by Nvidia. His company, Tenstorrent, was founded in 2016. Their Galaxy AI accelerator platform eschews the power-hungry High Bandwidth Memory (HBM) used by most competitors in favor of GDDR6 memory to improve power density by a claimed 3X over Nvidia systems, delivering almost 75 PetaFLOPS of FP8 performance in a single 60kW rack.

Inspired by how ants travel far beyond their homes but still find their way, researchers at the Delft University of Technology have developed a highly efficient method for long-range robotic navigation that uses very little computation or memory. The technology described in their research paper might be used in future tiny, autonomous robots that monitor warehouse stock or find gas leaks at industrial sites.

“Robotics is where A.I. meets reality”

Henrik Christensen, Chancellor’s Chair, Robotics Systems, UC San Diego

AI Insights

Like ice cream, artificial intelligence comes in a variety of flavors. It’s important to understand what business problems each flavor can solve so you can build your AI transformation strategy and put AI to work for your organization as effectively as possible. This Synthetic deep dive uses plain language to help you understand the difference between supervised learning, deep reinforcement learning, generative AI, and a wide range of other AI-related terms.

Early diagnosis of Alzheimer’s Disease (AD) is vital to improving patient outcomes. Early intervention is vital to slowing disease progression and extending quality of life. Over 55 million people worldwide are impacted by AD, yet until now, clinicians have lacked effective tools to accurately predict how early dementia diagnosis may progress in the future. A new predictive prognosis model, outlined in The Lancet, predicts disease progression and prognosis with 80% accuracy using MRI and patient cognitive test data.

Agentic AI, the next major frontier in artificial intelligence, offers the potential for a fundamental rethink of business models, moving from Software-as-a-Service (SaaS) to Managed-Services-as-Software (M-SaS). The on-premise model dominated in the ‘80s, ‘90s, and early ‘00s. For the last 15 years, the cloud has redefined software and service delivery. Does an entirely new business model present itself in the AI era? Perhaps so.

Kenza Layli has won the first-ever Miss AI contest. She’s a lifestyle influencer from Morocco and 100% created by AI. What she looks like and says are all AI-generated. Why bother with a Miss AI contest, you might ask? Beats us. Though the images of Kenza Layli and the words she/it utters reveal something about the state of the art of technology. Perhaps the most salient point is made in the last paragraph of the article.

Toolkit for the Future

AI-powered tools and services to accelerate your business.

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.

Need fresh talent that can harness the latest AI tools to increase the speed and volume of marketing content creation? Unlock value with fully-vetted, top marketing people with experience in your industry. The marketing-as-a-service model allows you to hire talent on demand, part-time or full-time, in 30-day blocks of time.

Need to expand internationally but unsure where to start? Oyster makes it easy to find, hire, and retain local talent. They handle the details (country-specific labor laws, international tax laws, compliance, and global payroll) so you can focus on finding the right talent and ramping their impact.

Focus on the email that matters. Deep Clean automatically identifies old, unimportant emails so you can delete them. Snooze emails until a better time and easily track emails that need a response. Automatically filter unimportant emails into specific folders. It works across every device where you check your email. Get to inbox zero with award-winning SaneBox.

Try this easy-to-use, AI-powered alternative to DocuSign to write proposals, create quotes, get eSignatures on contracts, build online forms, collect payments, and much more. Great for sales, HR, legal, finance, and IT. Free 14-day trial.

Create and sell online courses from your own website with this top-rated platform. Easy to use with hundreds of beautiful templates and a powerful generative AI assistant that makes content generation a breeze. The AI will create lessons, exams, assessments, and more. Free to try.

Relieve pressure on your support team and serve customers better and faster with a support automation platform. AI answers 90% of queries and escalates the rest to your team. It automatically updates its knowledge database so it’ll never escalate that same issue again. Supports SMS, social, email, and phone communication.