OpenAI It is reportedly preparing artificial intelligence (AI) models that will provide more advanced, human-like reasoning.
of The company’s o3 model aims to solve more complex, multi-step problems by spending more time computing answers before responding to user queries.Bloomberg News reported From the OpenAI livestream event on Friday (December 20th).
During the event, OpenAI provided details about how o3, an inference model announced in September, is better than o1. The company also hopes safety and security researchers will apply to test the model. before announcing new software, the report added.
OpenAI plans to release a smaller version of the model called o3-mini at the end of January, followed by the o3 model soon after.CEO Sam Altman said during the event.. A company spokesperson says there is no o2 model, “from A nod to the British telecommunications company of the same name.
As noted by Bloomberg, the announcement comes as other big AI companies jump on board. move forward with An increasingly advanced version of the model.
for example, google We recently published a new version of it Flagship model “Gemini”, It is said to be able to “think, remember, plan and even take action on your behalf” at twice the speed of past models. and Mehta recently hinted at his next plans put out the Rama 4 2025 model.
However, the report added: numbers One of these companies sees diminishing returns from costly efforts to develop new models. tactics such as make bigger concentration About “reasoning” help According to Bloomberg, they are avoiding the challenge of finding new, untapped sources of high-quality, synthetic data for training.
In other AI news, PYMNTS wrote about an AI system developed at MIT last week. warehouse robot Deftly handle oddly shaped loads and navigate crowded environments without endangering human workers.
nevertheless robot New PRoC3S technology excels at repetitive tasks such as moving pallets and could finally solve the long-standing problem of allowing robots to safely perform more complex warehouse tasks that require human dexterity and spatial awareness. there is.
“Theoretically, PRoC3S could reduce robot error rates by scrutinizing initial LLM (Large-Scale Language Model)-based assumptions against a more specific and accurate understanding of the warehouse environment. .” Eric NievesCEO and Co-Founder plus one roboticshe told PYMNTS.
“Think of it like this: It explains how a warehouse robot working solely on LLM guidance can complete a task. The concept of PRoC3S is to place a digital robot in a simulated environment for that task. , goes one step further. It’s essentially the difference between classroom instruction and school instruction. Really It was a good excursion. ”
