AI's Alien Intelligence Contrasts Human Experience, Not Replicates It

Key Quotes

"So much of the coverage about this stuff right now is like this running debate right where you've got people on one side saying these ai you know they think they are intelligent and eventually they'll outsmart and destroy us all and then on the other side you've got people being like no they they aren't actually intelligent they're just mimicking us and it's not as big a deal as everyone says and i i don't actually know who to believe and i think it's because like i don't know what ai is like i don't know how it does what it does under the hood because we don't know right this is one of the most extraordinary things about you know machine learning ai is that we don't really know what they are."

Latif Nasser explains that the public debate around AI is polarized, with some fearing AI's intelligence and others dismissing it. Nasser attributes this confusion to a lack of understanding about what AI truly is and how it functions internally, highlighting that even experts acknowledge this mystery.

"And so effectively these systems don't have the common sense of a mouse whereas higher reasoning math and so on they can do a hell of a lot better than humans can that's the that's the more of a paradox right like it's like easy things are hard and hard things are easy exactly and like we've known this for a long time and it's pretty obvious at this point but after running all of these ais through this thing dozens hundreds of times what steven has seen over and over is that they have a completely different profile of capabilities and skills than any animal."

Steven Cave's research, as described by Nasser, reveals a paradox in AI capabilities: they struggle with simple, common-sense tasks that animals perform easily, yet excel at complex reasoning and math beyond human capacity. Cave's repeated observations indicate that AI possesses a unique and fundamentally different set of skills compared to any animal.

"But then as it continued quizzing itself comparing its output to what it should have said when i go to my cousin's and i play badminton all that slowly we could actually hear the learning you could hear it figuring out the difference between vowels and consonants and then it would start pronouncing small words you know oh we please and uh you know it only took a couple of days when we walk home from school i like to go to my grandmother's house because she gives us candy and it was acing it and we eat those sometimes oh we play sometimes and then we sleep over there sometimes we sleep over there sometimes when i go to go to my cousins i get to play badminton all that."

Terry Synowski recounts the development of NETtalk, an early text-to-speech AI. Synowski details how the system, by repeatedly comparing its own pronunciation attempts to correct examples without explicit rules, gradually learned to distinguish sounds and pronounce words, demonstrating a form of experiential learning.

"And so effectively these systems don't have the common sense of a mouse whereas higher reasoning math and so on they can do a hell of a lot better than humans can that's the that's the more of a paradox right like it's like easy things are hard and hard things are easy exactly and like we've known this for a long time and it's pretty obvious at this point but after running all of these ais through this thing dozens hundreds of times what steven has seen over and over is that they have a completely different profile of capabilities and skills than any animal."

Grant Sanderson, explaining neural networks, likens them to a series of interconnected light bulbs representing pixels and processing layers. Sanderson describes how these networks are "wired" with random connections, and through a process of trial and error guided by mathematical feedback, these connections are adjusted to learn specific tasks, like recognizing a circle.

"And so what ibm was doing was giving it a bunch of texts books transcripts conversations feeding that into this machine and so then the right answer was the most likely word to follow the preceding words okay so it's like it's just like here's a giant stack of human talking and in this giant stack what's the most likely thing that would have been said next in this exact scenario exactly that's that's right and just one brief aside because it's sort of fun i i think i have this right that a word is a big long list of like 13 000 numbers what a computer has to turn a word one word just like a one word into 13 000 numbers yeah and so like in the way that a pixel value uh in the circle example was like basically a zero or a one it's like every word is this list of 13 000 numbers oh it's so weird that it like that that's that's the simpler version for it."

The explanation of early chatbots by an unnamed speaker highlights their predictive mechanism: by analyzing vast amounts of human text, they learn to predict the most probable next word in a sequence. This process, where words are converted into complex numerical representations, underscores the fundamentally alien nature of AI's internal processing compared to human language comprehension.

"And so with these gpus and this new parallelized architecture that google named a transformer all of a sudden they could get a machine to parse those longer sentences and give at least reasonable answers to more complicated questions all right but what really sent these ai chatbots into the stratosphere was a kind of knock on effect of this parallel processing because when you can process everything at the same time in parallel you can actually train on a lot more material in the same amount of time and so eventually they just gave it basically the entire internet almost everything we humans have ever said on the internet as its training material and started sending that through this network of light bulbs and wires that was just unimaginably big like to get a sense in our smaller example with the circle there's something like a thousand and some odd parameters right a thousand or so of those wires gpt 3 which was kind of dumb by today's standards but it came out had 175 billion parameters 175 billion things that could be tweaked yeah and many of the ones that we have now they're trillions of parameters."

The text explains that the development of GPUs and the Transformer architecture enabled AI to process longer sentences and larger datasets, leading to the training of models on nearly the entire internet. This massive scaling of data and parameters, as described by the speaker, resulted in AI systems exhibiting seemingly intelligent behaviors, a key factor in the proliferation of AI.