On Friday, during Day 12 of its "12 days of OpenAI," OpenAI CEO Sam Altman announced its latest AI "reasoning" models, o3 and o3-mini, which build upon the o1 models launched earlier this year. The company is not releasing them yet but will make these models available for public safety testing and research access today.
The models use what OpenAI calls "private chain of thought," where the model pauses to examine its internal dialog and plan ahead before responding, which you might call "simulated reasoning" (SR)—a form of AI that goes beyond basic large language models (LLMs).
The company named the model family "o3" instead of "o2" to avoid potential trademark conflicts with British telecom provider O2, according to The Information. During Friday's livestream, Altman acknowledged his company's naming foibles, saying, "In the grand tradition of OpenAI being really, truly bad at names, it'll be called o3."
It's been a really busy month for Google as it apparently endeavors to outshine OpenAI with a blitz of AI releases. On Thursday, Google dropped its latest party trick: Gemini 2.0 Flash Thinking Experimental, which is a new AI model that uses runtime "reasoning" techniques similar to OpenAI's o1 to achieve "deeper thinking" on problems fed into it.
The experimental model builds on Google's newly released Gemini 2.0 Flash and runs on its AI Studio platform, but early tests conducted by TechCrunch reporter Kyle Wiggers reveal accuracy issues with some basic tasks, such as incorrectly counting that the word "strawberry" contains two R's.
These so-called reasoning models differ from standard AI models by incorporating feedback loops of self-checking mechanisms, similar to techniques we first saw in early 2023 with hobbyist projects like "Baby AGI." The process requires more computing time, often adding extra seconds or minutes to response times. Companies have turned to reasoning models as traditional scaling methods at training time have been showing diminishing returns.
OpenAI is bringing o1, its “reasoning” AI model, to its API — but only for certain developers, to start. Starting Tuesday, o1 will begin rolling out to devs in OpenAI’s “tier 5” usage category, the company said. To qualify for tier 5, developers have to spend at least $1,000 with OpenAI and have an account […]
When AGI arrives, most won't even realize it, some AI experts say. Others say it's already here.
amperespy/Getty Images
Some say OpenAI's o1 models are close to artificial general intelligence.
o1 outperforms humans in certain tasks, especially in science, math, or coding.
Most people won't notice when AGI ultimately arrives, some AI experts say.
AI is advancing rapidly, but most people might not immediately notice its impact on their lives.
Take OpenAI's latest o1 models, which the company officially released on Thursday as part of its Shipmas campaign. OpenAI says these models are "designed to spend more time thinking before they respond."
Some say o1 shows how we might reach artificial general intelligence — a still theoretical form of AI that meets or surpasses human intelligence — without realizing it.
"Models like o1 suggest that people won't generally notice AGI-ish systems that are better than humans at most intellectual tasks, but which are not autonomous or self-directed," Wharton professor and AI expert Ethan Mollick wrote in a post on X. "Most folks don't have a lot of tasks that bump up against limits of human intelligence, so won't see it."
Artificial general intelligence has been broadly defined as anything between "god-like intelligence" and a more modest "machine that can do any task better than a human," Mollick wrote in a May post on his Substack, One Useful Thing.
He said that humans can better understand whether they're encountering AGI by breaking its development into tiers, in which the ultimate tier, Tier 1, is a machine capable of performing any task better than a human. Tier 2, or "Weak AGI," he wrote, are machines that outperform average human experts at all tasks in specific jobs — though no such systems currently exist. Tier 3, or "Artificial Focused Intelligence," is an AI that outperforms average human experts in specific, intellectually demanding tasks. While Tier 4, "Co-intelligence," is the result of humans and AI working together.
Some in the AI industry believe we've already reached AGI, even if we haven't realized it.
"In my opinion, we have already achieved AGI and it's even more clear with o1. We have not achieved 'better than any human at any task,' but what we have is 'better than most humans at most tasks,'" Vahid Kazemi, a member of OpenAI's technical staff, wrote in a post on X on Friday.
More conservative AI experts say o1 is just a step along the journey to AGI.
"The idea somehow which, you know, is popularized by science fiction and Hollywood that, you know, somehow somebody is going to discover the secret, the secret to AGI, or human-level AI, or AMI, whatever you want to call it. And then, you know, turn on a machine, and then we have AGI. That's just not going to happen," Meta's chief AI scientist, Yann LeCun, said on Lex Fridman's podcast in March. "It's not going to be an event. It's going to be gradual progress."
OpenAI released the full version of its o1 reasoning model on Thursday.
It says the o1 model, initially previewed in September, is now multimodal, faster, and more precise.
It was released as part of OpenAI's 12-day product and demo launch, dubbed "shipmas."
On Thursday, OpenAI released the full version of its hot new reasoning model as part of the company's 12-day sprint of product launches and demos.
The model, known as o1, was released in a preview mode in September. OpenAI CEO Sam Altman said during day one of the company's livestream that the latest version was more accurate, faster, and multimodal. Research scientists on the livestream said an internal evaluation indicated it made major mistakes about 34% less often than the o1 preview mode.
The model, which seems geared toward scientists, engineers, and coders, is designed to solve thorny problems. The researchers said it's the first model that OpenAI trained to "think" before it responds, meaning it tends to give more detailed and accurate responses than other AI helpers.
To demonstrate o1's multimodal abilities, they uploaded a photo of a hand-drawn system for a data center in space and asked the program to estimate the cooling-panel area required to operate it. After about 10 seconds, o1 produced what would appear to a layperson as a sophisticated essay rife with equations, ending with what was apparently the right answer.
The researchers think o1 should be useful in daily life, too. Whereas the preview version could think for a while if you merely said hi, the latest version is designed to respond faster to simpler queries. In Thursday's livestream, it was about 19 seconds faster than the old version at listing Roman emperors.
All eyes are on OpenAI's releases over the next week or so, amid a debate about how much more dramatically models like o1 can improve. Tech leaders are divided on this issue; some, like Marc Andreessen, argue that AI models aren't getting noticeably better and are converging to perform at roughly similar levels.
With its 12-day deluge of product news, dubbed "shipmas," OpenAI may be looking to quiet some critics while spreading awkward holiday cheer.
"It'll be a way to show you what we've been working on and a little holiday present from us," Altman said on Thursday.
On Thursday during a live demo as part of its "12 days of OpenAI" event, OpenAI announced a new tier of ChatGPT with higher usage limits for $200 a month and the full version of "o1," the full version of a so-called reasoning model the company debuted in September.
Unlike o1-preview, o1 can now process images as well as text (similar to GPT-4o), and it is reportedly much faster than o1-preview. In a demo question about a Roman emperor, o1 took 14 seconds for an answer, and 1 preview took 33 seconds. According to OpenAI, o1 makes major mistakes 34 percent less often than o1-preview, while "thinking" 50 percent faster. The model will also reportedly become even faster once deployment is finished transitioning the GPUs to the new model.
Whether the new ChatGPT Pro subscription will be worth the $200 a month fee isn't yet fully clear, but the company specified that users will have access to an even more capable version of o1 called "o1 Pro Mode" that will do even deeper reasoning searches and provide "more thinking power for more difficult problems" before answering.
OpenAI has launched a new subscription plan for ChatGPT, its AI-powered chatbot platform — and it’s very, very expensive. Confirming leaks this morning, OpenAI announced ChatGPT Pro, a new $200-per-month subscription tier that provides unlimited access to all of OpenAI’s models, including the full version of its o1 “reasoning” model. “We think the audience for ChatGPT Pro will […]
OpenAI’s “12 days of shipmas” event doesn’t kick off for a while, but the first big announcement might’ve been revealed early. A “feature gated” web page on OpenAI’s website refers to a “ChatGPT Pro” plan that includes all the benefits of ChatGPT Plus, as well as new goodies. Recall that ChatGPT Plus is an upgraded […]
A Chinese lab has unveiled what appears to be one of the first “reasoning” AI models to rival OpenAI’s o1. On Wednesday, DeepSeek, an AI research company funded by quantitative traders, released a preview of DeepSeek-R1, which the firm claims is a reasoning model competitive with o1. Unlike most models, reasoning models effectively fact-check themselves […]
