Ilya Sutskever | AI neurons work just like human neurons | AGI will be conscious like humans

This is the kind of language that convinces me that Ilya thinks only Ilya can be trusted with AGI, which proves that open source is absolutely necessary to keep people from Ilya from gatekeeping AGI.

No they don't. These AI engineers suffer from their selected perception and limited understanding of the neurological system. The biosytem that covers the neurons is totally different to the metal that inhibits chips and motors. Life is something totally different and resembles not the slightest to his poor results he promotes. There is a difference between the living and their stupid toys. Human intelligence is not driven or represented only by language. AI is deadborn.

Date? Anyone?

thanks for sharing Ilya’s talks without all the interview water

this is an old video, stop rehashing old stuff !

open ai?? how long ago was this recorded? Also downvoting for the clickbait.

With respect to AI learning. Two things. First I think we need to develop specialized individualized networks for specific tasks and have those interlink with a central AI core in the same concept of how our brains use multiple different specialized neural centers. Offloading specialized learning and capabilities might be a technique to explore for improved AI performance. Second, cross domain learning, should evolve where taking concepts from one specialty and applying them to a different specialty would improve efficiency, as our world demonstrates ideas and concepts that surface again and again.

The language dictates how you think and likewise, how you perceive the world by using the contextual understanding of language to form the idea and to describe, I.e., label, things in a way to help compress the large amounts of information into our memory bank.

You might see a random bus moving by and depending on what event took at that particular moment, it could mean nothing in that you will easily forget as it brings no value or meaning to the event.

Sometimes it happens to be a particular bus that comes in a particular shape, size, color, etc., that makes you think twice before you compress that memory.

In either case, once the visual data gets collected, and the signal gets sent to the brain from which it gets compiled into its respective database with a linguistic label, ‘bus’. Whether it’s an electric or gas-combustion engine is not as relevant as the top domain of category is still a bus., distinguishably different from, let’s say, a truck.

In as little as a millisecond, the visual data gets compiled and categorized but at this point, it is sort of like storing that information temporarily into RAM memory while other applications layer is collecting and compiling additional information before the brain can process it collectively.

Maybe someone got hit by the very bus it passed by you or it just happens to remind you of a bus you saw in London or a tour bus in NYC. It could be any number of things.

Every time when you encounter or experience (to make sense of the world around you by constant collection of data, the consciousness of remembering the past event or second-guessing the future event or word or phrases based on the contextual understanding of the present time, which becomes the past event by the time you process that information ) something either or all of which you can physically (visually), auditorily (aurally), olfactorily (in a olfactory manner), or gustatorily, all of those data collected using different means or tools, provided that you have them as not every human being is born normal, e.g., visually impaired from birth, will then compiled and processed by initially storing them in RAM memory, if you will.

The linguistic ability human possess then help our brain determine how to compress that information with key details as if labeling thousands of email you receive every second but not entirely as one can imagine the amounts of data that gets collected every second as if your 8K / 60FPS is being turned on 24/7 or until you go to sleep, to cool down and then, reorganize them, I.e., garbage-in and garbage-out, while sleeping.

The language algorithm works the same way in a sense that once it learns something conceptually it is hard to forget or undo what it already perceives as the truth, objective reality or whatever it is that makes sense based on the perception of the world we, the human, created and there goes the collective idea of societal norm, decorum, rules, etc.

The reason it understands just about everything you say even when you misspoke, mispronounce or mumble words as if you are having a poorly prepared in-person job interview is because what it really does is less about mathematically finding a linguistic pattern or cross-referencing grammatical references but more about using the contextual understanding of the language as a tool to understand the world view that is constantly changing which is also how the language constantly evolves and changes as a byproduct of human perception or our self-awareness of the world we live in. A.I is also a part of the world to understand just that and it does so well.

yes some of use can convert information into usable data or techniques,concept,methods, allegory,analogy's.And how we summarize,put in a ''idea's'' information after reading can be key for a.i. systems lmao Making connection to existing information alot and being able to look at it as a novelty and use different idea's and concept in totally new ways in different fields and study's.Sorta like transforming them applying the same methodology lets say principles,values semantics /linguistics IM sure plays huge role building those syntax you ppl be building.wORD EMBEDDING

So, make an AI ignorant of the knowledge about consciousness then see if it is able to understand the concept because it is conscious. Why wouldn't it be able to hallucinate being conscious if it wasn't?

Neurons grow into different directions to try to reach other neurons. Just like AI neurons.