Listen to Steve read this post (7 min audio) – those who listen get extra commentary – I go off piste!
There’s a huge debate going around right now on whether or not Artificial Intelligence is actually ‘intelligent.’ While some say it poses a threat to our species, others argue that it is far less advanced than anyone realizes. Included in the latter cohort is AI expert Rodney Brooks, whom I believe to be among the best in his field. He contends that we’ve been vastly overestimating OpenAI’s large language models, asserting they are much ‘stupider’ than we assume. While that may well be true, it doesn’t matter. Let me explain why…”
Check out my 2 min video on p(Doom) – be sure to comment & send it to a friend.
Brooks asserts that large language model AIs are simply adept at generating what an answer should look like, which is distinct from knowing what the answer should be. When asked if AI is poised to become the sort of artificial general intelligence (AGI) that could operate on an intellectual level similar to humans, he responded:
“No, because it doesn’t have any underlying model of the world; it doesn’t have any connection to the world. It is correlation between language.“
Intelligence doesn’t need to be ‘Human’
Brooks’s error lies in his analysis from a solely human perspective. His view is that for intelligence to surpass us, it must behave like us. It must possess an internal sense of knowledge, a context of the world. However, even in nature, we observe a considerable number of intelligent species with entirely different contexts from humans. Biological beings go about achieving their objectives and expanding their species using various forms of intelligence. There’s no reason to believe that AI can’t or won’t do the same. More importantly, AI doesn’t need to understand in order to act. I don’t comprehend how my heart beats, how my digestive system functions, or how I can control my arms to catch a ball, yet I can perform these actions. The questions we really need to ask are:
(A) What is intelligence?
(B) Can it exist in non-biological forms?
Bonus – Listen to me be Interviewed by Sir Bob Geldof on the risks of AI.
While there are many types of intelligence – abstraction, knowledge, understanding, self-awareness, learning, emotions, reasoning, planning, creativity, critical thinking, and problem-solving – we can more generally define it as follows:
Intelligence: The ability to absorb data, infer information, retain it as knowledge, and then apply it to generate outputs relevant and useful to a particular context.
Through this definition, we could consider plants and ecosystems as intelligent, and clearly, AI. Understanding is not part of the equation. For all we know, this could be a purely human phenomenon, and irrelevant to creating new forms of intelligence outside of the biological realm.
If AI begins to organise information and the physical world independently, and without direction from humans, then what it does or doesn’t know about itself or the world doesn’t matter. If it is acting as an intelligent species, it will impact other species – and become part of the wider ecosystem.
While AI is simply a pattern recognition device, so too are humans. Large Language Models just being able to predict what ‘sounds right’ might actually be enough. Especially when the data input is language – because language is the single mechanism that holds together all forms of human knowledge. Hence, they do not need to have a direct connection to the world – they instead have one which lives inside a different context – a computational one. And it is therefore foreseeable that these models will be able to predict and see patterns we cannot, and make decisions in their context.
This is some feedback I got from a client on my my new Keynote on AI –
“It was best keynote I have seen in 20 years – it was that insightful.”
It’s a revolution – Get me in – Don’t miss out!
If you take anything away from this post – let it be this:
Don’t let yesterday’s definitions of anything dictate what your business, your future, or our world might look like.
(Oh, and have the courage to disagree with others in high places if you’ve done thorough study via reputable sources.)