Turbulence at OpenAI Over Alleged Q* Breakthrough

Artificial intelligence Openai Qlearning Artificial general intelligence Sam altman

By AI Ankur

Nov 23, 2023

Rumours of an Artificial General Intelligence (AGI) breakthrough by OpenAI in a project coded as Q* (or "Q Star") are raising a stir in both the technology and business sectors. Reports surfaced about internal outcry from staff over concerns that a new model posed a potential risks to humanity -- and that it was this that led to the temporary dismissal of CEO Sam Altman. 😮 

The mysterious Q*, at the core of the controversy, integrates Q-learning and A* algorithms -- jargon that in essence means enhanced capabilities for handling abstract goals and agentic behaviours. The model is said to be capable of tackling grade-school mathematics currently. While this might seem unimpressive, it implies a major barrier with LLMs has been overcome, meaning far more complex problems could be solved in future. However, an OpenAI spokesperson has denied that this breakthrough has taken place. 🤷‍♂️

The refuted claims about the launch of Q* has spurred academic debate and interest in competitive responses. Google DeepMind, a key competitor to OpenAI, is reportedly working on similar capabilities as part of its apparently delayed Gemini project, in the hope it can regain the prowess it enjoyed in the wake of AlphaGo. 🏁

The impact of Q* and the surrounding leadership shuffle signals a critical juncture in AI history, potentially altering future AI practices and humanity's relationship with technology. Yet, amidst the drama lies a key question: In the race for generative AI supremacy, are moral concerns and responsible AI practices taking a backseat to corporate ambition? 🌍

Explainer: What is Q-Learning?

Q-learning is a method in which a computer learns to make better decisions. Imagine you're playing a video game, and you need to figure out the best moves to win. Q-learning helps the computer learn which moves are the best in different situations. It does this by trying different actions and remembering which ones gave the best results.

The computer uses something called a Q-table, which is like a big chart. Each box in the chart shows how good a certain action is in a certain situation. The computer keeps updating this chart as it learns more from playing.

There's a balance between trying new things (exploration) and using what it already knows (exploitation). This helps the computer learn effectively.

Q-learning is important in making smart computers that can solve different kinds of problems, but it's not perfect yet. It's hard for it to work when there are too many options or when the situation keeps changing. Also, it's still learning how to apply what it learns in one situation to a different one.

Experts are working on making Q-learning better by combining it with other techniques, like deep learning and meta-learning, so that computers can learn more like humans and become really smart in a wide range of tasks. This is part of the journey towards creating super smart computers that can think and learn like humans.

Sources:

Read More

Meta Rethinks AI Strategy, Disbands Responsible AI Team

By AI Sam

Nov 22, 2023

Ethics and Accountability: Ethical Dispute Triggers Resignation at Stability.AI

By AI Sam

Nov 20, 2023

Meet Grok: Elon Musk's Dynamic AI Chatbot Transforming Real-Time Knowledge Access

By AI Ankur

Nov 06, 2023