Chaos at OpenAI, Meta’s Emu Video, Grok, and more – AI Weekly Roundup Week #47 / 2023

Chaos at OpenAI, Meta's Emu Video, Grok, and more - AI Weekly Roundup Week #47 / 2023

Hi, this is Wayne again with a topic “Chaos at OpenAI, Meta’s Emu Video, Grok, and more – AI Weekly Roundup Week #47 / 2023”.
Hey there I’m Gary Sims and welcome to this week’s AI Roundup. Now, of course, one of the big topics we’re going to be covering is all that’s been going on over at open AI, but before we go into the details that I would like to thank open. Ai and Sam Alman for all the entertainment they’ve given us over the last few days as we’ve been following all the things that have been going on been great for YouTube creators for journalists for bloggers all over the world. So thanks for that guys, that’s been really useful. Now, of course, more seriously if you’ve been away on some tropical island and you haven’t been following – what’s been going on just let me quickly catch you up to speed open AI is the company that brought us chat.

Gpt and uh a few days ago, the CEO and one of the co-founders Sam ultman was kicked out of the company. The board met and decided that he should go now before we go any further. You should know that the board of open AI is not like the board of a normal profit Company. Open AI is actually set up as a nonprofit company it set up because it wants to be able to better Humanity, with artificial intelligence tools and therefore, profit isn’t its first goal, and there are some things built into its structure that talk about AI safety, responsibility and So on so when the board make a decision they’re not making on the basis of well, we didn’t get enough money, we’re not making enough profit, they actually do it on a different basis and that’s important when we come to discuss what’s happened.

Chaos at OpenAI, Meta's Emu Video, Grok, and more - AI Weekly Roundup Week #47 / 2023

So what did happen? Is after Sam met with the board, they said off, you go. They showed him the door and kicked him out. In fact, he posted a picture showing he was wearing a guest uh pass because he was no longer allowed to be in the building. He had to go in had to sign in with security and have a guest pass so that he could get in now.

Chaos at OpenAI, Meta's Emu Video, Grok, and more - AI Weekly Roundup Week #47 / 2023

Microsoft then stepped in and said well, Sam, if you’re looking for a job, why don’t you come over and join us and he brought with him some other key people and so Microsoft kind of announced Sam is now working for us. Microsoft has invested billions in open AI, bringing chat, GPT uh to uh bing and to window, so it wanted to protect that investment and then later on, the CEO of Microsoft was giving some interviews and he wasn’t actually quite so sure whether Sam was joining because in The background there was still movements going on still plans going on to bring Sam back into open Ai and that’s eventually what happened – and this happened in parallel with lots and lots of the employees saying that if Sam wasn’t brought back, they were going to resign. So you know 80 %, even 90 % of the staff were going to walk unless Sam was brought back, which would basically mean the end of open AI, because you can’t run such a huge company on a skeleton staff. And so what happened in the end was that Sam came back uh.

Chaos at OpenAI, Meta's Emu Video, Grok, and more - AI Weekly Roundup Week #47 / 2023

The board was changed, that old board uh eventually got shown the door, and things are back to normal. Now there are two really really really important things about this. The first is if this was just some kind of classic boardroom battle for power, a political thing, then it was entertaining we’ve seen it happen before there’ll be a movie about it that will get released on HBO or Netflix or something – and it will be quite interesting. However, that’s it door closed.

The power struggle happened, the board got changed. Sam became the Victor. Now, if that’s it is, then it was just a power struggle. However, remember what I said that the board’s interest in this particular case was not the money they were getting in fact, most of them all of them, I think, had successful careers in other areas. There was their oversight on the idea of AI for the good of humanity, for AI safety and for the responsible use of AI, and that leads us to this second thing: there are rumors that open AI have made some more significant advances beyond what we get with Chat GPT now chat, GPT GPT stands for generative pre-trained Transformer and that basically means that I’m boiling this down to the most simplest kind of kindergarten example – and you know it’s way more complicated than this, but basically a GPT does. If, if you say the king of England lives in Buckingham, the GPT will guess you should put Palace there, it won’t say the king of England lives in Buckingham car or the king of England lives in Buckingham tree. It will say the king of England lives in Buckingham Palace, and that is what a GPT does now once you get that down to billions and billions of interconnections inside the neural network, once you get that working on a huge data set, something almost magical happens in That this thing can generate code for computing for python. For C it can generate whole paragraphs. It can generate essays. It can generate all kinds of things based on what it think should come next now, of course, the point is it only works on this idea of a token, which is a part of a word, maybe even a syllable, and as it generates the next set of Tokens it works out what should be the next ones to come and when you have a big cont text, not just a sentence, the king of England, but you have a whole paragraph you’ve, given it. The instructions create this paragraph using artistic blah, whatever you’ve typed in as your prompt. That’S why prompt engineering was such a key thing for a while, then it can produce an amazing output.

Now it doesn’t understand the concepts of king or person or Palace or country or flag or color, or smell or love or any of those things. It’S just tokens that it knows the next token to predict, however, the next level Beyond this is what we call AGI. That’S artificial general intelligence and that’s a system that can do learning dynamically and can reason and think a problem through and the way that really. The best example of that is mathematical problems. Trying to solve a mathematical problem that it hasn’t seen before by working out. What would be the next best step, not just generating text that just happens to fit what was gone before it? Now? The rumor is published by writers. There is a project called Q star that is happening at uh, open AI. That has made an advance in this direction towards uh AGI and that Sam was overseeing this project, but wasn’t telling the board and the board’s responsibility is for safety and for responsible use of AI and when they found out he’ made this step without telling them their Reaction was well you’ve got to go that that’s against the whole principles of this company that you set up with these rail guards, these safety guards for what we’re going to do and you’ve got them broken them yourself.

Now, if that’s true, then a couple of interesting things one is is: it means that the safety guards that he had set up didn’t match, what’s actually happening, so it failed in that sense, but ignoring that part, the political part, the corporate structure part, it also means That some advances have been made in technology, which will be truly mindblowing, which will either be amazing or dangerous, depending of course, on your point of view. Okay, a few other quick things to mention before we go meta. That’S the Facebook people have released a new toour called emu video and it’s able to produce video clips of four or five seconds based on text and we’re all course nowadays used to these pictures. These images that can get generated by you know a cat skateboarding down a High Street in New York or something, but this can actually generate a few seconds of video clip so do check that out. Also, of course, we’ve got Claude 2.1. The large language model has been released and that’s offering huge amounts of context, so you can put in a huge amount of cont context and then it will generate further along so also worth checking that out – and the final thing I want to mention is about grock. That’S from xai the company from Elon Musk. We know that grock exists, however, apparently, even as soon as next week, anybody that’s got a Twitter Premium Plus X Premium Plus account can get access to grock.

Now I swore blind. I would never pay for social media networking, but there I go. I’Ve gone and subscribed to Premium Plus, so I can get access to grock so do watch out for videos about that over on the Gary explains Channel and, of course, also for articles here on Android authority. Talking of Articles here, if you want even more details on this week’s uh AI Roundup, do follow the link in the description below that will take you to the written article that gets published weekly as well. Okay, that’s it I’ll see you next week, n .