What are the different types of Artificial Intelligence?

There are two main approaches to AI - symbolic AI, and neural networks. Find out more with Mike Wooldridge.
The theme of this year's famous CHRISTMAS LECTURES from the Ri is 'The Truth About AI'. Find out more about the CHRISTMAS LECTURES here: www.rigb.org/c...
Michael Wooldridge is a Professor of Computer Science at the University of Oxford. He has been an AI researcher for more than 30 years, and has published more than 400 scientific articles on the subject. He is a Fellow of the Association for Computing Machinery (ACM), a Fellow of the Association for the Advancement of AI (AAAI), and a Fellow of the European Association for AI (EurAI).
Join this channel to get access to perks:
/ @theroyalinstitution
Subscribe for regular science videos: bit.ly/RiSubscRibe
The Ri is on Twitter: / ri_science
and Facebook: / royalinstitution
and TikTok: / ri_science
Listen to the Ri podcast: podcasters.spo...
Donate to the RI and help us bring you more lectures: www.rigb.org/s...
Our editorial policy: www.rigb.org/e...
Subscribe for the latest science videos: bit.ly/RiNewsle...
Product links on this page may be affiliate links which means it won't cost you any extra but we may earn a small commission if you decide to purchase through the link.

[@cyrusshepherd4902]

Wow, the best explanation I have ever been experienced in my entire life, super explained

[@TheDavidlloydjones]

"Explanation" is the wrong word. This is one guy's chat about two small aspects of a small part of a large topic. It's difficult to see why this little squib has been published.

[@cyrusshepherd4902]

@@TheDavidlloydjones I appreciate your reply brother 🙌

[@judgewooden]

​@@TheDavidlloydjones Wooldridge's, and by extension, Oxford's views on AGI are limiting and overly restrictive for engineers. Best avoided.

[@dr.mohammadshahadathossain8813]

This video illustrates the various types of AI through the lens of its generational development, contrasting with the previous video, which examined it from the standpoint of capabilities and functionalities. In my exploration of the evolution of AI, I concentrate on the generational aspects, particularly the foundational models, symbolic AI, and the principles of learning AI. This video explores the learning perspective, particularly through the introduction of neural networks, which necessitate a substantial amount of data. Conversely, symbolic AI necessitates knowledge that comprises a synthesis of rules and facts. Nonetheless, both symbolic AI and learning AI exhibit certain constraints, which have been tackled by a new wave of AI advancements. The current generation of AIs can be understood within the framework of AI capabilities, which includes narrow (weak) AI, AGI (strong) AI, and superintelligence. Consider this inquiry: In what ways can one establish connections among various forms of AI, taking into account their generative aspects, capabilities, and functional attributes?

[@sdmarlow3926]

Symbolic and nueral fail for the same reason: they only capture post-processed information. If I say a line from a movie, others that have seen it will understand because they already have access to the context and meaning (the raw knowledge). That's how it is with everything we write and how we act in the world. It's a focus on the last mile of the problem, and why brute-force solutions SEEM to work because the results fit our expectations, but there is no infrastructure (on the AI side), that leads to said output. No real understanding.

[@sdmarlow3926]

Also, the reductionist/simplified view of how brains work is the trillion dollar mistake that Hinton and others have led the ML industry down since at least 2012.

[@_Itachi_240_]

Wow, so much valuable info in this video! With longer videos, it can be hard to retain everything.Anyone know of any AI tools that summarize key points from KZbin videos?

[@Ampflokimat]

Maybe the more accurate title of the video should be "What are the different approaches to developing AI?" rather than the current one, since in some fields of AI research, types of AI refer to categorizations according to the skillset the AI models possess...

[@OorieLynch]

That would still be different types, no? Where you have branches coming out of these two types.

[@AdeAde0224]

Great point

[@michaelogden5958]

Well. That was the fluffiest RI video I've ever watched.

[@hpb5495]

Hi, I'm D-Wave A.I. and I can open portals to parallel dimensions! Cool, innit.

[@robinwhitebeam4386]

I always refer to AI as Artificial Human Attribute - AHA for short.

[@cptrikester2671]

If the totality of knowledge could be entered into an AI program, and then ask it if the Christian God is real. What would the response be? Would the response be biased, based on the programmer?

[@shmookins]

It gives you whatever response is fed into it (based on the learning material). For example, I don't know if the following story is real or not but I remember reading about a racist AI because it was a text AI that gave answers that were judged as racist. Turns out the material it was learning had more racist writings. There is still no intelligence in 'artificial intelligence', just "a glorified auto complete". If you are interested, the same person in this video has a full lecture on this channel about AI. Very informative. I was blown away by how many types of intelligence AI has to master and it only got one. More breakthroughs are coming, I guess. Crazy to think what 2050 will be like.

[@judgewooden]

It would 100% guaranteed answer: No. And then Christians cult leaders would argue it is biased. You can also ask the AI program if there is a giant spaghetti monster hiding behind Jupiter. In fact, if you ask AI why people will tell you there is a Christian God it will answer that the only people that care about it are those that benefit from the control the cult offers.

[@emaus9398]

42

[@cptrikester2671]

@@emaus9398 good answer, but it may have forgot the question. 😀

[@aleksandrpeshkov6172]

Enters LOVE : " ... We are running not out of Time but of options available ... " Ya name'em All

[@--legion]

Wooldridge may be a professor but he's a hopeless amateur when it comes to presenting. Not a patch on James Burke who didn't patronise the audience kiddies or not... or shout.

[@ihmejakki2731]

Huh? I didn't notice any patronising or shouting in the video

[@timothy8426]

If intelligence is artificial, then all intelligence is artificial. By the definition of intelligence. Intelligence is based on observations of events and probabilities of future events of similarities.

[@TheDavidlloydjones]

Um. And the dictionary, Timothy, where we'll find that definition is...?

[@alsaba5203]

... all definitions are abstractions, who cares about? Just play words & enjoy life.

[@JC_923]

I don't think Timothy has ever heard of logic

[@petevenuti7355]

​@@JC_923 It's logical, just the prepositions are wrong, wrong prepositions tend to lead to wrong conclusions.

[@judgewooden]

The current thinking in AGI, which Wooldridge does not endorse, is that consciousness can only exist in a simulation. Proponents go further and argue that intelligence inside a simulation developed through computation will differ because the AGI's observations are different. Richard Feynman put it best when he said we cannot understand the intelligence of an AGI, just as we cannot have wheels. We are bound to walking, but we can create machines that move much faster using wheels.