Depending on who you ask, artificial general intelligence might not be the looming game-changer some companies have made it out to be.
What happened: Nick Frosst, co-founder of Canadian AI company Cohere, said the money being thrown at developing artificial general intelligence (AGI) doesn’t “make a ton of sense.”
- As a scientist, Frosst believes developers and companies need to be honest about what AI is actually capable of.
- As a business founder, Frosst is doubtful that AI’s capabilities will scale by pursuing AGI, and continuing to invest in it doesn’t seem like a sustainable business model.
What it is: Artificial general intelligence is an AI that can match humans on cognitive tasks, whether that be in terms of knowledge level or the ability to learn and reason. This can let it do more complicated, multi-step tasks, unlike today’s generative AI which needs a bit more hand-holding and very clear instructions.
- AGI would be particularly useful for things like robotics, to help machines react to situations in industrial settings with less human supervision.
- Or, it could be put to use in AI agents that can think about the best way to complete a task — for example, finding a recipe and placing an online grocery order for the ingredients when asked to help pull together a healthy dinner.
Catch-up: Miles Brundage, who left his post as a senior advisor on AGI at OpenAI last week, wrote that no company, or the world at large, is ready for AGI. Brundage is on board with AGI as a concept but is concerned that no one has developed proper governance to use the technology safely or manage its broader impacts.
Why it matters: These are some of the first comments from people actually working at high levels in the industry to cast healthy skepticism on AGI, something the likes of OpenAI, Google, and Amazon have been eagerly pursuing in hopes it will expand their customer bases so their large investments can start generating revenue.
Bottom line: While (most) executives have been optimistic about AGI, researchers in the field have been doubtful about whether or not it will be developed any time soon, or if it’s even possible.