Hi, this is Wayne again with a topic “GPT-4 is already old news”.
I need to turn to the washroom. It’S my mid-show pea Break um, see you later uh see ya, um time to talk about AI. What’S new, Mr Luke got him. What do you have for us uh? I haven’t. It’S honestly been very hard to keep up with everything, because it’s literally been moving that fast at this point in time, gpt4 itself is like almost old news, uh, which is kind of crazy. I don’t like that.
A lot yeah, I’m not okay with much. I think there’s very legitimate reasons to not like that. I don’t.
I don’t want to talk to you about it on stream, but yeah every week you give me a panic attack, let’s go um, so we had a question in here about that. The craziest thing, in my opinion, I think this happened this week. Um, there’s I mean there’s been updates to the repo as of five hours ago, so it’s very live just like everything in this space is, but Microsoft announced uh, something called Jarvis, and this is just one of the actually many options of this type of thing that I’Ve seen announced or talked about recently – and this is where things are getting really crazy. I talked I think a couple weeks ago.
I think this is going to be an absolutely massive jump when these things hit more uh wide adoption, because plugins with chat GPT. Okay, I’m gon na I’m gon na rewind, slightly make sure everybody’s with me. Before I keep going so chatgpt extremely powerful Limited in ability right.
It can write creative things really really well, it can’t do things for you so much in the coding space. It can do things for you to a certain degree. There’S some limitations there. It will literally spit out code that has errors stuff like that, but you can work with it to fix those errors or you can just work. Uh use it as a very good like starting off point for small projects stuff like that.
It has ability to do things in the coding space, but it can’t really do things in other spaces and it has issues with math and all this other type of stuff. Well, a lot of those problems were fixed very quickly, but in a private sense, because not that many people, including myself, I’m very sad about this – not that many people have access to plugins plugins was announced. I think a couple weeks ago, and the first plug-in that immediately I was like, oh my God. If they make this, this is going to be absolutely insane and it was literally like the first one they did was Wolfram Alpha Wolfram Alpha is super cool, has been super cool for a long time and is very not new um.
I think people that aren’t in the space haven’t necessarily heard of it, but Wolfram Alpha is an answer engine developed by Wolfram research. It answers factual queries based by Computing uh. Sorry, it answers factual queries by Computing answers from externally sourced data. Wolfram also Alpha was released on May 18, 2009.
yeah and the the the type of things that it does is like quite advanced, math and calculus. So you take chat GPT, which can’t really do these things and you give it a plug-in where chat GPT can ask Wolfram Alpha for the answer and suddenly wow, it’s extremely powerful right now. The next stage of this is not just plug-ins but Jarvis, and I think they call it hugging face which is like that’s a concerning name but Jarvis and hugging face, and all this other type of stuff allows you to link all these systems together. So you could ask chat GPT, you could be like hey.
I need to do this thing and chat GPT or any other llm can reach out, find the different services that it thinks it could use to accomplish. That task assign various tasks to the different Services, get those services to actually execute those tasks, get responses from those Services potentially iterate, multiple times. If it needs to and then spit out the result to you, so it can do more things now. It can actually get things done.
That’S a big big step. Um, there’s also a bunch of other things that happened, uh someone’s suing. I think it’s open AI for defamation, because um yeah it like made up something about them.
It called them a criminal or something when they were actually a whistleblower yeah, pretty cool, that’s going to be a huge issue. I mean it’s wrong all the time all right. So if you start asking it about people, it’s going to also be wrong about people yep and that’s going to cause some problems.
Um yeah, there’s, there’s there’s lots of stuff going on. In my opinion, if you just kind of like dabbling in this stuff and only looking into it a little bit, I would look into like Microsoft’s Jarvis um hugging face stuff like that. In my opinion, that’s where the most interesting stuff is happening um this second and I think that’s where I’ll kind of leave it isn’t hugging face more like an online repo of all these AI Tunes yeah, but you use Jarvis to get an llm to tap into Hugging face to find the correct one: oh good God, no yeah, no, no, no, no yeah, yeah, yeah yeah. All right continue! Linus! Please did you talk about the Samsung semiconductor leak in your AI update, no just Samsung employees um using chat GPT to fix coding errors, resulting in the leak of test sequences for identifying chip defects as well as source code of a top secret application. I told you I told you this stuff was gon na happen. That’S hilarious.
I’Ve been told by so many people like. Oh no people aren’t going to do that. I told you this is gon na happen. Oh my goodness, I’m so not surprised! This is why the whole data Silo thing is going to happen. Mainframe yeah, I don’t think so. I don’t think you have okay, fine talk about your AI Theory.
Let’S go I’ll. Try to keep it short. Let’S go. I I’ve had a a pretty good track record. I think literally 100, so far with making certain calls about AI stuff, I’m making a much more long-range and likely to be an hour out there yeah, but I think, like Mainframe style systems like big, intense lots of companies have them less cloud-based kind of stuff is Potentially, coming back hold on David R, it’s Apple leather made of apples not like from Apple. My reason for that is stuff like this, because I think a the productivity tool that is um like what I’m going to call Mesh AI.
I don’t know um llms, also working with other AI systems or or other various tools, Etc is going to be extremely valuable. It takes a massive, very intense amount of compute to run these systems. So not only is there going to be a lot of compute needed, but the security behind it is going to be extremely extremely high. So I am certain.
Oh also you. These things will work better if they ground themselves in your data, and you want these things to be able to have access to like your entire accounting history and all of your inventory and all of your production information and basically everything about your company so that they Can help your employees that work on those things uh to to be able to work with them right like we saw in Microsoft’s example that it was going through campaign data for marketing and telling you how well it worked in different ways and all this kind of Stuff based on it being in Excel and whatnot yeah, that might be fine for some places, but on-prem solutions for AIS going to be a thing. I am so certain of this for certain really ultra high security applications. It probably already is for some applications for some super high security, stuff yeah, where you’re like having the cloud solutions for this is not an option, but we’re going to hit a point where not using AI for productivity reasons. Stupid is also just dumb.
So that’s not an option yeah, so you have to do it so you’re gon na have these big, crazy, on-prem things you’re gon na have like Microsoft is going to make a killing. Oh yeah, it’s gon na be crazy, yeah. So there’s going to be solutions that are in the cloud that are for normal users, there’s going to be solutions for Enterprise companies that are a little bit less concerned about the data that they put up there, but they still need it to be secure. So those will be cloud-based and then there will be like hey we’re, uh government entity of some kind or a hospital, maybe or I don’t know a bank or Microsoft or something crazy, so we’re going to have our own local one that is grounded in our data. It’S used for us, it’s it’s changed in different ways and then it’s going to get really weird, because AIS are going to learn to improve themselves. So all these different things that are siled in their own areas are going to improve in their own directions and things are going to get crazy.
That’S my that’s my long! That’S my long shot! I think that’s gon na happen and there’s stuff, there’s stuff. That literally came out this week, even more things like there’s, Lama and alpaca. Already there was a new one. I don’t remember the name of it, a new release that says that it can compete with modern GPT that runs locally. I don’t remember what it’s called, but it runs locally, so they’re like it, it doesn’t call out whatever so it can be secure. You can also tie those into the voice control systems that I mentioned a while ago, so you could use like the Siri API through a through a Macbook. That’S why for it to speak to you and for you to be able to speak to it and that’s running locally, and while I still can’t dictate who I want my phone, this stuff is literally more effective than that already well yeah. I should hope so this should be more effective than that. It should because it should just check a local index of things and compare against it, because that’s like 10 years ago technology anyway, someone mentioned my last comment: someone mentioned AI Hardware, specialist job.
I think there’s going to be a job which is like in in my brain. I see it as like a tender or like an arborist, but instead of that, it’s for managing and shaping and helping add some like human direction to the self-growth and self-improvement that AIS are going to have. I think that’s going to be a role making sure you’re you’re continuing it to go in the direction that the company wants it to go in prompt director yeah.
I don’t know yeah .