OpenAI and Meta will release AI models capable of reasoning and planning – a key step towards "superhuman cognition".
Although AI has come a long way in the past few years, there are still some areas where machines can't do as well as humans.
However, that could all change soon, as both OpenAI and Meta are reported to be about to release AI with reasoning and planning capabilities.
Leaders of both companies say that the latest version of the AI model is coming out and will be even more powerful.
According to their report, ChatGPT-5 and Llama-3 will not only generate text, but also begin to engage in activities more akin to thinking.
Joelle Pineau, vice president at Meta, said: "We're trying to figure out how to make these models not only conversational, but also inferred, planned, ...... Have memories. ”
Meta has told that the company will soon begin sharing early versions of its Llama 3 model, along with releasing a comprehensive series of models, which are scheduled for release in 2024.
OpenAI executives told the Financial Times that they plan to release GPT-5 "soon."
Currently, apps like ChatGPT use artificial intelligence called large language models (LLMs) to predict which words should appear next in a sentence.
By being trained on large amounts of data, AI has become so good at predicting strings of words that it is starting to sound like an intelligent, thinking agent.
However, Meta's chief AI scientist, Yann LeCun, said that even the best AI can still "make stupid mistakes."
Speaking at a Meta event in London on Tuesday, Mr LeCun said that the current AI "really doesn't think and plan, just produces words one after another".
He added that adding reasoning capabilities would allow AI to search for possible answers, plan actions, and create "mental models of the impact on its actions."
Brad Lightcap, COO of OpenAI, said: "I think we're just starting to tap into the inference power that these models have.
We're going to start to see AI handle more complex tasks in a more sophisticated way.
I think over time...... We'll see models move toward longer, more complex tasks, which inherently require them to be more inferentially proficient. ”
At the same time, Mr. LeCun said that Meta is working on a new AI "agent" that will be able to plan and book every step of the journey from Paris to New York.
These changes will be a significant advance in the capabilities of AI models, and perhaps even one step closer to creating artificial general intelligence (AGI).
AGI is seen by many as the ultimate goal of AI research, aiming to build machine learning systems that can reach the level of human cognition on any task.
To achieve this, AI needs to be able to complete complex task sequences and predict their actions.
While not all researchers believe this is feasible, Elon Musk recently predicted that by the end of next year, AI will become smarter than the smartest humans.