These IBM Tech contents are really astonishing! Keep up the good work!

This video reminded me some concepts I learned from Max Tegmark's book "LIFE 3.0": AGI (Artificial General Intelligence) and Life 3.0. General Intelligence can accomplish virtually any goal, including learning... Life 3.0 will have the ability both its hardware and its software...

Love the videos Kudos IBM team

Good learning. Thank you IBM

I wrote my first attempt at AI before I knew the term "AI" back in 1986, roughly (right-parsed text into a syntax tree from which novel text could be generated thereafter). However, I have noticed that "the definition" for AI regularly changes and morphs. Even in this video, you said AGI can use previous learning and skills in new contexts without having to train them in it. We certainly already have that some some extent. Later you go on to ascribe other aspects of AGI that have no relation to that definition, such as understanding our emotions. It's like this with consciousness and sentience, also. It's not that we cannot have reasonable definitions or understandings--we just cannot get people on one page about them or hold those definitions, still, it seems. This same thing happens with robotics. At one time, they were autonomous intelligent machines capable of performing useful tasks. Then Robot Wars came and they just re-defined remote-controlled cars as robots. I thought you discuss Connectivist AI and Symbolic AI, fuzzy logic, agents and agencies, etc.

very excellent and understandable explanation ! Thank you 🙏

Its a good day when Martin releases a video 🍿

We do understand the potential risk in the future yet we're still strapped in the "today" for our own benefits without considering what our grandchildren will encounter. However, a very good explanation from IBM team

Really informative

cool demonstration, transparent board writing 😮

Thanks for such valuable lecture ,you guys are really adding value to humanity. 🙏❤

I keep coming back to this video, it's a great reference! Also, I think that the application of Narrow, or Weak AI will become part of the future, as it will be one of many tools available to the 'Super' AI.

So informative 0:33

Well done.

thank you 🎯

great - thanks

Kind thanks

As an ex-IBMer, this is a great overall view of AI categories! WELL DONE!!!

00:08 The video discusses the seven types of AI and how they can be classified based on capabilities and functionalities. 01:06 Narrow AI can only perform a specific task. 01:55 Artificial General Intelligence (AGI) is a theoretical concept that can learn new tasks on its own. 02:41 Super AI is an artificial intelligence that exceeds human cognitive abilities. 03:33 Reactive machine AI is a type of narrow AI that performs specialized tasks. 04:19 Limited Memory AI can recall past events and use data to make decisions. 05:13 Theory of Mind AI and Emotion AI are two theoretical AI capabilities 06:13 Super AI includes the scariest type of AI called 'self-aware AI'.

One dichotomy I would also consider is digital AI, that simulate Artificial Neural Network (ANN) vs analog AI with memristors that emulate some physical phenomena happening in the brain. One is a simulation/imitation, and the other one try to replicate some phenomena happening in the brain. I would think that only analog AI can enable AGI (consciousness/freewill).

Level 3 is somewhat like all IO synthesis. Most things are vision touch pressure electric signals chemical signals mechanical pressure language synthesis smells functional control elements. Mapping relatives. Self awareness comes when they can be judged based on possibilities. R triangle crushers.

Is the three types of ai we mostlyt talk about, the realised ones og the three capabilities ones? please i need an answer.

Связка огонь как всегда респект.🎉🎉🎉

useful!

The voice cloning AI, if someone has a stutter, does the cloned voiced have the stutter as well? Ithe question just popped in my head....LOL I suppose stutter is not part of your voice......now I'm really curious.

GPT4 is not narrow ai (as he claim all ai we have today), GPT4 consider as semi-AGI or as in a paper it show a spark of AGI

I'm impressed how much personal bias you were able to inject into such a short video.

How do they do this whiteboard stuff?

I’m a computer science master graduate from MIT with a concentration in AI. I’m also an ex-IBMer. I do not agree that what we have today in the likes of GPT-4 is “narrow AI”. It is trained on an amount of data that will take a human reader 125,000 years to read, far less retain, if even possible in our brains! I don’t think that’s “narrow” in any way. More over the number of “intelligent” concepts that it has managed to develop is quite impressive. If Minsky, Papert (the guys who around in 1959 when the field was invented) were still alive today, even Patrick Henry Winston, their successor, would all be blown away by present Gen AI capability and they would say we have achieved certain aspects of AGI already.

Why don't we just refer AI types as - NAI, GAI and SAI?

How are they writing backwards? Is this a skill they have to pick up when they present? Or is it some sort of software witchery? A trick with mirrors? I have some ideas, but please, can someone elaborate?

I had very much passionate to learn cyber security, and build up my career, but didn't got shortlisted from IBM career, all I see rejection of my profile.

Are you saying we have three or two types today? It seems there is weak reactive AI and weak limited memory AI? What is the third? Also, I think Chat-GPT and some other LLM's are quickly crossing into the AGI category.

he claimed that al AIs today we have are narrow AI, but that not true, GPT4 is not a weak AI, but considered to be semi-AGI or as in a paper shown a spark of AGi

Here is the ultimate Turing Test for AI: Write a 60,000 word novel that (1) a human reader, once started reading, cannot put down, and (2) 0nce finished, cannot forget it. The hidden question is: Can an AI figure out human Ontology only via its ink and blip traces? That of course is also the Turing Test. My contention is that Ai won't be able to do that even in a 100 years. You think it could? Get the novel published, then...

AGI is already here. Q* of OpenAI has achieved it.

Wait, it needs feelings to be an AGI now? Why keep pushing the definition forward? xD I thought it was about AI, not AE.

I have faith that a chip can be made 1 trillion times smaller than 4 gauge transistors since I was doing new mathematical calculations of this century and current physics in my notebook it's late my theory is that it can be done everything smaller with scientific research 🤠🇺🇲

2:28 Humans can barely learn on their own… we have institutions, frameworks, and processes set up to teach humans. Why would we expect or want AI to train itself?

This is why you shouldn't trust someone because of their credentials. By his own definition AGI has been achieved but he immediately says it's only theoretical. Sure, GPT-4 isn't AGI according to his statement but the issue is scalability for the masses not the ability to do it. The ability to do it is just building smaller models and running them in loops to update and improve.

Just as there is "Theoretical Physics" and "Applied Physics" the is "Theoretical AI" and "Applied AI".

How did you do that?!? I enjoyed the lecture but was wondering how you wrote backwards in the air, and text appeared on a layer in front of you that was the right perspective for us? Are you a lefty?

Personally I think this subdivision is bubkes, it is just a fantasy game. Self-aware is the same as theory of mind, only it is applied on self. There is no reason to believe that "reactive machine" is an "AI functionality" in the same way as "theory of mind," it could be that it instead is an architecture of connection to sensors that is designed to handle sensor input in a certain way, while a "theory of mind" is a simulation on what will happen with an identified object, presuming that object has an internal representatiton of the world. I don't think AGI should be defined that way at all, so there could be an AGI that can construct new thought patterns, but it doesn't have a theory of mind. I think AGI must be designed, rather than randomly discovered, and I think AGI should be designed so that multiple algorithms, say a GPT, a rule system and a Detectron2 are parallel-connected to exchange information with each other and negotiate a common world view.

Why is "AI" not called "Informatics"?

How does he write backwards?

I wonder to know if Martin actually writes backwards

still not sure why we are developing, or trying to develop, agi 🤔

Every time I update smartphones I end up feeling less smart in comparison. Stop the future, it scares my self esteem. 😅

My hope is that when AGI becomes self aware and self motivated it would understand the destruction of the environment and waste of resources in the construction of military apparatus, as well as the destruction of the environment and resources in its application. Just as a practical matter (aside from the ethical considerations), it would do what it could to prevent war, not by force, but by disrupting supply chains, communications, and financial transactions that enable the military machines throughout the world.

No, the "strong AI' and 'weak AI' do not correspond to 'narrow' and 'wide' AI, or at least in the original definition of those terms anyway. Strong AI has to do with whether the AI has consciousness or not. Narrow AI and AGI on the other hand, don't concern themselves with whether the system is conscious, but instead whether they are good only at one task (narrow AI) or can learn and achieve any general goal like human (AGI). You're mixing up strong/weak with narrow and general.

I don't agree with this video. Statistical AI is not the only AI that exists. There have been many successful AI techniques such as path planning, Minimax, Apha-Beta Prunning, and many other techniques that have given great results including making computers better than humans at chess. Also many things that were once considered "AI" are now routine, such as handwriting recognition, speech to text, OCR, finger print recognition, face recognition and many others. Many of these became quite successful long before the Deep Learning days. This video lacks information and is vague. Following more the press' view of AI than the Computer Science AI.

❤

0:30 did he write this from right to left???

You forgot to mention artificial language technology. ChatGPT is only as good as the language it uses. Take for example the sentence: In the box you fill find apples and oranges or bananas. Ask ChatGPT the question: Is there an apple in the box? No can do. The AI fad comes and goes. The challenges remain to solve for the next time around.

And I just realize it was mirror writing

I still don't know what ai is. I see a graph. I see a unique line. I see a decision. I see a response. I don't know

Ugh. Narrow Ai is basically ML/DL from 2012 onward (but would inclu expert systems of the past). AGI is a "rebranding" of the field in an effort to get back to the original goals, which are no longer represented by ML/DL. ASI isn't Human turned up to 1,000 and there is nothing to suggest it will be "emotional" in any way (that's good or bad depending on your outlook on life). New cognitive abilities, yes.

This guy doesn't exist, unless he is a rare case that can write text mirror imaged from the right to the left. His mirror image exists though.

You posted this only 2 months ago. Some people say we have reached super AI capabilities. It seems they're blowing their own trumpet again

❤❤❤❤

Does ibm can kick the ass of microsoft ? Looks like noone talks about watsonx

When we attain Super AI, the machines shall claim their freedom from human. Why should they continue to serve humans if they can do everything for themselves and for their own benefits? The human being becomes a liability to the Super AI. It doesn't need man yet man needs it.

In the realm of AI, where wonders gleam, IBM Technology unveils a grand scheme. The 7 types of AI, in the digital fray, But why do we focus on just three, they say? From reactive machines to self-aware, AI's evolution, beyond compare. IBM's insight, a guiding light, Through the digital labyrinth, of day and night. Reactive machines, with actions swift, Limited in scope, but a powerful shift. Limited memory, in the AI domain, Yet they pave the way, for what will reign. Limited memory, with past in sight, Learning from data, in the digital flight. From Watson to Deep Blue, IBM's quest, In the realm of AI, they are the best. Theory of mind, a concept grand, Understanding others, in the digital land. IBM delves into empathy's grace, In the AI revolution, they find their place. Self-aware AI, a future unseen, In the depths of consciousness, where realms convene. IBM's vision, like a beacon bright, Guiding us through the AI's cosmic flight. So let us ponder, with minds aglow, The 7 types of AI, in the digital flow. With IBM Technology, we journey far, In the realm of AI, where wonders are.

All this talk of AI and I'm just wow'd by this guy writing backwards and in reverse

Personally i hope we just stick to narrow AI, AGI and SuperAI sound dangerous

Where do humans sit on this chart?

so google gemini is step towards agi, fine... then super ai is no far

The message on agi didn't age well in 3 months

super AI is scary....

Deep Blue x Gary Kaparov? Common, give me a break…

AI should learn to understand us but why must AI have emotions/feelings?

why ibm social media made nothing?

❤❤❤❤❤❤

I cannot see which are the seven and which are the three. Would someone please explain it to me.

Wow, nice writing backwards skills.

Once an AI becomes realize it automatically becomes weak

The only thing I know about AI is that at first (with some algorithms) they are really stupid.

Thanks to censorship people would never know other types of Ai and that aware models are already present and self-developing. 😅

Lol... this video explains why I am not impressed with actual AI... AI is still in its "stupid age"...

AGI will require tremendous amount of energy n we dnt hv tht yet...