The Limits of AI: Generative AI, NLP, AGI, & What’s Next?

[00:00] Artificial intelligence is everywhere right 

[00:05] emails for you. You may be wondering if there 

[00:11] heard many people over the last few decades 

[00:16] but it's never going to be able to do, and 

[00:20] of those predictions have in common? They were 

[00:25] growth in AI capabilities, bringing it from the 

[00:30] most of those things that so many thought 

[00:35] many limitations still exist, but my advice would 

[00:42] you want to be wrong. In this video, we're going 

[00:47] how it differs from data and information, and 

[00:52] a look at what have been considered to be the 

[00:58] have actually been accomplished and what's still 

[01:02] about the role of AI and humans and where each one 

[01:09] amazing technology to our best advantage. Let's 

[01:15] exists among data, information, knowledge, and 

[01:23] out. So, we'll start with data. Okay, this is just 

[01:30] like this, I say 10 six uh 42 and 8. Okay, that's 

[01:39] do with that, but that's data for you. Okay, now 

[01:47] information. So this is where we sort of processed 

[01:53] that this data actually represents the ages of 

[02:01] This has more meaning to us. Now if I take that 

[02:08] to the information that we just had. Now we end up 

[02:16] So for instance in this case we might say okay 

[02:22] room are under the age of 21. So now we've done 

[02:30] last piece of this is applied knowledge. Applied 

[02:38] look at this all of this information, all of this 

[02:44] what, we've got these people in a room. Let's 

[02:50] to keep them occupied. So, uh the 42-year-old 

[02:57] a 10-year-old and a and an 8-year-old would play, 

[03:01] for a little while. So this is an example very 

[03:06] here data information knowledge and wisdom each 

[03:14] and all of these then lead to the ultimate of 

[03:20] is data. Well, that's a database. For instance, 

[03:25] but that's all it is, just a collection of raw 

[03:30] running on a computer. That's now information 

[03:35] added context to all of that data. Knowledge, this 

[03:41] adding more interpretation to the information 

[03:46] still trying to get. And that's wisdom. Back when 

[03:52] studying AI in its earliest days, there were a lot 

[03:57] of AI. Maybe one day we'll have a system that's 

[04:02] maybe even in our lifetimes." For instance, one of 

[04:07] to reason. We needed a system. If we really 

[04:12] is a part of that. So the ability to figure out 

[04:19] uh this was beyond our capability uh certainly in 

[04:26] computer that can play chess. IBM in 1997 came 

[04:33] Gary Kasparov, the best chess player in the world, 

[04:38] a lot of problem solving. People thought you'd 

[04:42] a grandmaster. Again, that's already happened. So, 

[04:49] that was really difficult for a long time was 

[04:56] has a lot of nuance, a lot of idioms, things 

[05:01] And sometimes you're supposed to interpret it 

[05:06] Uh for instance, as I've given examples before, if 

[05:10] doesn't mean that there are small animals falling 

[05:16] system is going to really be intelligent in 

[05:20] understand those things. It needs to be able 

[05:24] when you're cracking a joke and when you're not. 

[05:28] and sometimes it's because it's a bad dad joke. 

[05:34] tell the difference between what is humor and what 

[05:39] here. In 1965, there came about a first the first 

[05:47] was not using modern technology called Eliza. And 

[05:53] it wasn't very great conversations, but it would 

[05:59] how are you feeling today? How does that make you 

[06:04] like you're talking to one of these very passive 

[06:11] 2011 when we came out with Watson which played 

[06:18] and beat champions at that because Jeopardy is 

[06:24] and things like that. You can't program all of 

[06:29] really has to understand the meanings behind 

[06:34] as I say, we've already accomplished that. And 

[06:39] to understand a lot of this nuance and they're 

[06:44] natural language and understand what you mean in 

[06:50] one of the most remarkable aspects of generative 

[06:55] the first time. We feel like a computer really 

[07:01] asking for. In some cases, even anticipate the 

[07:05] would. We consider that to be intelligent. 

[07:10] I remember hearing a lot of people say, you know, 

[07:16] they actually do. Uh, we've got where with 

[07:22] new works of music. And you can say, well, but 

[07:27] guess what? When people compose a new song or draw 

[07:34] we've heard as well. Listen to all the top musical 

[07:39] yeah. Here are my musical influences." So, those 

[07:43] influenced the way that they create. So, we are 

[07:48] the old. But that doesn't mean just because a 

[07:53] fact it is. They're coming up with new ideas and 

[07:58] our creativity on certain things that have 

[08:04] here's another one. Real time perception. 

[08:12] of science fiction at one point, but we have them 

[08:16] but a self-driving car is one of those where it's 

[08:22] see what's going on, anticipate where the next 

[08:28] be at a specific point in time and do all of 

[08:34] uh decisions about that. Robots are having to do 

[08:39] So all of these things that basically we used to 

[08:46] you know what, we've done all of those. Now, let's 

[08:51] progress, but I don't know if we would say, you 

[08:55] one of those would be uh the area of you've 

[09:02] intelligence uh and an index for that? Well, these 

[09:10] I feel like some people are just able to 

[09:13] but that's a whole other subject. But an EQ in 

[09:19] the ability for them to understand your moods 

[09:24] So there is some level of awareness in terms of 

[09:28] we have the stories about people who felt an 

[09:34] some people feel emotional relationship to their 

[09:39] that these systems can talk to us and understand 

[09:45] moods and things like that is certainly in the 

[09:51] at least in some cases. Now, another area that's a 

[09:57] of hallucinations. Hallucinations are a difficult 

[10:03] AI where the system basically confidently asserts 

[10:10] predict what the right answer would be and many 

[10:15] But when it's wrong, it is shockingly wrong in 

[10:20] are making hallucinations less and less likely. 

[10:26] uh helps with this where we feed additional 

[10:31] model doesn't just use its own imagination to come 

[10:37] helps as well where we have different models 

[10:42] Uh so there are things that we can do in order 

[10:49] doing that. So this is one of those I wouldn't say 

[10:55] that we're moving into it. So this one's somewhat 

[11:01] kind of already done or are still working on 

[11:06] move those out of the way. And now, let's take a 

[11:11] current limits? What are the problems that we're 

[11:16] of the limits of AI is a thing called artificial 

[11:22] are super smart in a specific area, in a specific 

[11:28] bots that we have today, they seem to know a lot 

[11:33] limitations. For instance, they don't do real-time 

[11:38] for instance. So artificial general intelligence 

[11:43] doing all the things that we consider to be 

[11:48] person would do across all the different domains. 

[11:52] achieved in a single system yet. The next level 

[11:58] where we have something that is better than 

[12:03] now again the stuff of science fiction. Not saying 

[12:08] yet. Another problem that's still to be solved is 

[12:16] that can do amazing stuff but boy do they suck 

[12:22] They need lots of cooling. They're very expensive 

[12:27] able to scale if we just keep throwing more and 

[12:32] going to work. We're going to end up using all the 

[12:37] uh to to run some of these queries. So, we're 

[12:42] decisions with sustainability. Use models that are 

[12:47] the right size model. In some cases, a small model 

[12:52] might even hallucinate less if we've got the right 

[12:58] we're doing that is not yet, I would say, a solved 

[13:02] about it. Another one that is really the area that 

[13:10] is a system self-aware? Does it know it exists? 

[13:17] know the answer to that. This is really not a 

[13:22] question. So, I'm not going to try to deal with 

[13:27] answer would be. But another thing that gets us 

[13:33] a system can spit out a lot of things, but 

[13:39] does it really know what the meaning of the 

[13:44] but there's always the question of is this really 

[13:50] I don't know. I'll tell you there's a lot 

[13:53] may be only simulating thought and simulating 

[13:59] draw the line clearly, but uh this seems to be an 

[14:06] the biggest broadest context that we'd like it 

[14:12] was talking about data, information, knowledge 

[14:18] the business of wisdom and judgment. And in this 

[14:26] Maybe ethical judgments. Can it determine what 

[14:32] that? Some people have a real hard problem with 

[14:37] to program a system that will if we can't figure 

[14:42] know that right now these are limitations that 

[14:47] something that's just very subjective like the 

[14:52] what I think is really great music, you may 

[14:56] you have no judgment at all." Uh, but I have a 

[15:02] able they're able to generate music and they're 

[15:06] gibberish but can they tell what is going to be a 

[15:11] music area. So there's a lot of of work here in 

[15:16] qualitative judgments as well. How about this one 

[15:24] in uh in quotes because air quotes because um I 

[15:31] again we have limitations with people. So we can't 

[15:37] what we consider to be common sense because we all 

[15:42] there are some things that we know and the systems 

[15:48] there are some certainly some limitations to 

[15:53] some people would say that with today's agentic 

[16:00] go off and accomplish those things. And what I'm 

[16:05] micro goals. These are sort of the small things 

[16:11] and the macro goals. So the larger task, this 

[16:18] doing it. And right now today's agents are able 

[16:23] the larger objective, but the big goal, why would 

[16:29] uh without uh beyond its reach at the moment. And 

[16:36] AI system really sense things? Does it understand 

[16:42] taste? Um that sort of stuff. The things that 

[16:46] that are able to certainly see and hear. Can 

[16:52] but there's a lot of other things that go into uh 

[16:58] together in one system. And then here's the really 

[17:05] system really able to feel the same way that we 

[17:12] experience sadness, loss, uh, accomplishment? Does 

[17:19] I know some people who don't really do all 

[17:24] of the things that is difficult to put into a 

[17:31] really feeling these kinds of things? So, I would 

[17:35] to one degree or another are limitations with 

[17:41] and for AI? How do we work together? How do we 

[17:47] people really should be over here doing this kind 

[17:53] we want to do? That's the overall macrolevel goal, 

[18:00] What's the purpose of this? Is there meaning in 

[18:05] we're trying to accomplish? And without purpose, 

[18:11] are still far better at that kind of thing. And we 

[18:17] Over here on this side, once we've told the system 

[18:23] agent figure out the how and go off and perform 

[18:30] lot of things much faster than a person could. and 

[18:36] to know what to do in the first place. We need 

[18:41] it felt like for the longest time we were making 

[18:47] just took off. And we're at this this inflection 

[18:53] is going to go, no one really knows. But what I 

[19:00] of milestones that we've accomplished already. 

[19:06] that still need to be done, which is actually 

[19:10] it and the possibilities of problem solving, then 

[19:16] ultimately we're going to get end up with systems 

[19:22] So my advice to you if you start looking at 

[19:28] become preoccupied with those because the people 

[19:33] this that or the other thing have generally been