Building a Superintelligence: AI vs Human-Machine Hybdrids

Показать описание

Hey guys, trying something a little different today! This is a video discussion on AI, exo cortices and the singularity. To summarise: it is often suggested that we are heading toward the singularity - a point in time when technological advancement will accelerate at an unprecedented rate - and that the catalyst for that will be artificial intelligence. I suggest that actually, AI is a long way from being capable of that - but that a man-machine hybrid might be closer. This would use an 'exo-cortex' - an external machine capable of enhancing human cognition. This would allow us to think in whole new ways, resulting in exponential technological an scientific development.

Over the course of this discussion, I go into depth regarding topics such as AI, machine learning, virtual reality, embodied cognition and more and I provide a little crash-course and some food for thought regarding the current extent of artificial intelligence.

Like I say, it's a bit different so please let me know in the comments if you appreciate this kind of content, or if I should save this heady discussion for the blog.

In general, I'd love to hear your thoughts on all this. Thanks so much for watching.

Рекомендации по теме

Комментарии

Hey guys! Did you find this interesting? Would you like to see more transhumanism and discussion on this channel? And with regards to the topic of this video, what do you think the future holds for AI and for man-machine interfaces? Let me know and thanks for watching! :-D

AdamSinicki

One of the most interesting videos I've watched - never got bored and clicked away. I was amazed to see afterwords that this only had 361 views, i really thought it would be more in the range of 361k. Very professional and stimulating, please keep these kind of videos up man, you're doing a great job.

TheApologeticWeasel

I found this video interesting and refreshing, sharing a subject whos matter to you is a good idea, because you can do it with enthusiam and enthusiam is contagious. Most youtuber does not goes beyond the "boundaries" to what they channels are "labeled", to me you showing you are versatile, so keep up the good work.

DeSpaceFairy

I'm waiting for the time when you can 1-click-buy your Exo-Cortex from Amazon

SiimLand

Keep talking about the singularity
Everyone needs to be informed about this... plus I need to know what my next upgrade should be

markusdemerius

Great stuff Adam, as usual, and I would love to see more of this content :)

Some thoughts:

(1) The human brain is a very specially-evolved object which, while clearly able to engage in general-purpose reasoning, might not respond well to arbitrary upgrades. Executive memory, for example, is not well understood, and its role in general cognitive function is perhaps even less well understood. It's not known whether we'll be able to just plug RAM sticks into our frontal lobes and get better working memory -- and even if we did, it's not clear that the rest of the brain would be able to adjust.

(2) Getting a squishy, wet thing like a brain to play nicely with electronics is its own problem, and a difficult one at that. Even if we had solved (1), solving (2) might take a good bit of time.

(3) You're conflating 'consciousness' and 'intelligence' here. In my reading of the literature care is taken to distinguish between these two things. Making a 'superintelligent' machine may or may not necessitate making it conscious. And in fact a disturbingly compelling case can be made for consciousness being a hinderance to intelligence (note that most of what we call 'practicing' is a deliberate attempt to internalize a skill well enough to obviate the need of being conscious of the process involved). For more, see Peter Watts' outstanding 'Blindsight'.

(4) The most promising model I know for human consciousness is the notion that it arises from recursive isomorphisms embedded within the neural networks of the human brain. Above a certain complexity threshold the structure of the brain is able to generate a model of itself that is embedded in the causal stream, and this is what gives rise to the experience of consciousness. You cite consciousness as being the ability to make unpredictable choices, but even here an advanced machine intelligence could accomplish this by deliberately behaving randomly. In fact it is sometimes game-theoretically optimal to do just that, in order to make observations of one's behavior less useful in predicting future behavior (see: 'Artificial Intelligence: A Modern Approach')

(5) Concerns about dangerous superintelligent AI usually don't hinge on the possibility of any human programmers actually making an entity smarter than themselves. Rather, the concern is that a recursively self-improving AI would bootstrap itself well beyond human levels, possibly in a short amount of time.

(6) Nice summary of the simulation argument.

(7) There are fairly compelling exocortices right now. Check out Fastcompany's "A Head for Detail", and Steven Berliner Johnson's discussion of the DevonThink tool.

trentfowler

Would it make the reconstruced clip from brain activity more clear by using stuff like photo enhancing programs that make blurry photos more focused or something like facial recognition but for more than just faces? Or using software like the Hubble telescope does to enhance the most distant images it sees? Or an I completely wrong?

freerunner

The only purpose is to pass the butter

zonimorzine

Please more videos like this. I will love the Bruce Lee video when it comes out.

PhenxRsng

Please make more videos on this channel :)

AliMohebali.

Hi Adam, I find these none traditional videos on your channel interesting but incredibly long. I find it hard to agree that we would invent AI that would wipe us out. There are a number of TED talks on the topic of AI that summarise it better but any AI we invent would be bound by us as people. It would be amazing if these types of videos were more concise.

rbarnard