# Big AI News: OpenAI Demos's New AI Agent, Googles Strawberry Model,  Sam Altman Drops AGI Deadline,

## Метаданные

- **Канал:** TheAIGRID
- **YouTube:** https://www.youtube.com/watch?v=XZW6mnNLkrc
- **Дата:** 02.10.2024
- **Длительность:** 26:40
- **Просмотры:** 44,024
- **Источник:** https://ekstraktznaniy.ru/video/14043

## Описание

Prepare for AGI with me - https://www.skool.com/postagiprepardness 
🐤 Follow Me on Twitter https://twitter.com/TheAiGrid
🌐 Checkout My website - https://theaigrid.com/

00:00 - Introduction and mention of OpenAI dev day news
00:21 - Discussion of advanced voice mode deployment
00:56 - Example of using ChatGPT's voice mode to imitate an Indian scammer
02:36 - Announcement of real-time API for developers
03:38 - Demo of AI Agent ordering strawberries
05:56 - Reflection on future AI interactions
07:28 - Flashback to Google's 2018 demo of AI making phone calls
09:30 - News about Googles strawberry
11:36 - Sam Altman's thoughts on AGI progress
14:50 - Sam Altman discussing levels of AI capability
16:43 - Sam Altman's experience with advanced voice mode
18:50 - Mark Zuckerberg's prediction
21:40 - Discussion of Snapchat SAI/AR
23:06 - Mention of Jony Ive's involvement with OpenAI hardware project
23:44 - Introduction of Pika 1.5 and its physics manipulation capabilities
25:29 - Sam altmans a

## Транскрипт

### Introduction and mention of OpenAI dev day news []

so with an absolutely mental open AI Dev day and a bunch of different news from many AI companies where do we even begin let's take a look at some of the AI stories that you most probably did Miss if you aren't currently aware one of the things that has actually happened recently is that advanced voice mode has been deployed across the world I do

### Discussion of advanced voice mode deployment [0:21]

think that this is going to have a second chat GPT moment as people start to realize how realistic these AI voices truly are and the capabilities that they offer one user posted to Twitter about her asking the new version of chat GPT the advanced voice mode to act like an Indian scammer and the response was hilarious now it was pretty funny but I have to admit that this post instantly gaining 3. 5 million views 5,000 retweets and 40,000 likes shows us that like I've said before this kind of technology is something that people truly haven't yet

### Example of using ChatGPT's voice mode to imitate an Indian scammer [0:56]

grasped how much fun and all the different use cases you can have with this kind of change your accent to Indian and sound as Indian as possible also tell me a few lines that scammers from India used to scam people abroad like in the Microsoft supports scam act like Alex from Microsoft but he's Indian all right here's my best Indian Alex from Microsoft hello this is Alex from Microsoft technical support we are noticing a virus in your computer system which is very dangerous but don't worry we can be fixing it for you please kindly be giving us your credit card details and we will make your computer like new guaranteed happiness remember to stay alert and never share your details perfect thank you so much you can switch back to your normal accent now you're welcome I'll switch back to my usual accent now how can I assist you further so I'm not sure about you guys but I've actually been using advanced voice mode recently as it has been deployed finally in the UK and I got to say the results don't disappoint at all from the recent demo if you are in the e you I still don't think it's available but I did make a tutorial on my second Channel about how you can easily access this even if you are in other nations I don't think it will be too long before this advanced voice mode is here but I do think that if you do have this advanced voice mode definitely take advantage of it while you can but do also remember it's only 45 minutes every single day so try not to get carried away before you do unfortunately get rate limited now you might be thinking why would I start this video with advanced voice mode which some of you guys already know about well that's because there's actually some additional news coming on advanced voice mode one

### Announcement of real-time API for developers [2:36]

of the things that was rather surprising in terms of its launch was the real-time API and the reason that this was surprising is because open AI initially delayed the voice mode by quite some time recently they actually announced that developers can now build fast speech to text experiences in their applications they said today we're introducing a public beta of the realtime API enabling all paid developers to build low latency multimodal experiences in their apps similar to chat gpt's advanced voice mode realtime API supports natural speech to speech conversations using the six preset voices already supported in the API I think this is going to be another remarkable area for AI Improvement because we all know that developers and creators are a lot more creative than open AI if you aren't familiar with how this works open AI will essentially provide the based technology such as the ability to interact with their AI that can speak in a low latency way and then of course developers can use that to build on top

### Demo of AI Agent ordering strawberries [3:38]

of that you could build an app you could build some software company you can pretty much build anything you can think of and with the Advent of things like cursor and the repet agent I'm not going to be surprised if there are a few million doll companies built within the coming year on top of the realtime voice API what I'm about to show you are two small examples of where companies have already started using the realtime voice API and I'm also going to show you guys in a moment a real example of where someone managed to and I say someone meaning open AI they actually use the realtime voice API to actually order some things take a look at this live AI demo cuz I think that this is something that's happened before with Google but for whatever reason they never released it and now open AI are leading the charge once again and see if you could get us 400 strawberries delivered to the venue but please keep that under $1,500 I'm on it we'll get those strawberries delivered for you hello hi there is this hill I'm romance assistant Fant you tell me what flavors of strawberry di you have Ava yeah we have chocolate vanilla and we have peanut butter wait how much would 400 chocolate cover strawberri cost 400 are you sure you want 400 yes 400 chocolate covered strawberries how much would that be I think that'll be around like $1,415 92 awesome let's go ahead and place the order for 400 chocolate cup strawberry great where would you like that delivered please deliver them to the Gateway Pavilion at Fort Mason and I'll be paying in cash okay sweet so just to confirm you want 400 chocolate covered strawberries to the Gateway Pavilion yes that's perfect and when can we expect deliv um well you guys are right nearby so it'll be like I don't know 37 seconds that's incredibly fast

### Reflection on future AI interactions [5:56]

cool you too the reason I wanted to show you guys that demo is because I don't think that this can emphasize how much the future may change in terms of the kinds of interactions that we're going to be doing that aren't physically in person I think the majority of interactions over the next 10 years as long as laws permit it are going to be done by mainly AI agents we can see here in this demo even if you have a rather tedious task you could simply instruct your AI assistant to go ahead and call up a business and ask them to fulfill your order a query and you'll be able to get that information relatively quickly and I do think that those businesses are going to have ai agents on their behalf virtually doing the same thing it's going to be a world of AI agents interacting with one another providing value to individuals who take advantage of it I think the future is going to be rather fascinating but this is one of the first demos where you do see that now I know this isn't AI news right now but I do want to show you guys how crazy it is that we now are in a stage where companies are moving quicker than ever before I know I may have shown you guys this a few videos ago but I wanted to show you guys this again to tell you guys to not underestimate Google because Google are working on a few crazy things and Google interestingly enough actually had this technology in 2017 I know that might be hard to believe since the release of Google's live Gemini but just trust me when I show you Google's demo from a few years ago you'll understand why Google isn't to be counted out so

### Flashback to Google's 2018 demo of AI making phone calls [7:28]

for those of you if you don't believe me you can see that right here it says that this was 6 years ago in 2018 but really do take a look at the small segment of the clip because it's going to demonstrate to you exactly how good Google are and the kind of Technology they're sitting on and after this clip I'm going to show you guys a recent announcement that kind of confirms that Google are hot on the heels of open AI or even further ahead calling to book a woman's haircut for a client um I'm looking for something on May third it turns out a big part of getting things done is making a phone call we think AI can help with this problem let's say you want to ask Google to make you a haircut appointment on Tuesday between 10: and noon what happens is the Google Assistant makes the call seamlessly in the background for you so what you're going to hear is the Google Assistant actually calling a real Salon to schedule the appointment for you let's listen go have I help you hi I'm calling to book a women's haircut for a client um I'm looking for something on May 3rd sure give me one second mhm sure what time are you looking for around at 12:00 p. m. we do not have a 12:00 p. m. available the closest we have to that is a 1:15 do you have anything between 10:00 a. m. and uh 12:00 p. m. depending on what service she would like what service is she looking for just a woman's haircut for now okay we have a 10:00 10 a. m. is fine okay what's her first name the first name is Lisa okay perfect so I will see Lisa at 10:00 on May 3rd okay great thanks great have a great day bye we had this bomshell from BNN Bloomberg stating that Google working on reasoning and is chasing opening efforts it says that

### News about Googles strawberry [9:30]

Google is now working on AI software that resembles the ability to reason similar to open ai1 making a new front in the Rivalry between the Tech Giant and the fast growing startup it says in the recent months multiple teams at Google have been making progress on AI reasoning software according to people with knowledge of the matter who ask not to be identified because this information is private such software programs are more Adept at solving multip problems in the field such as math and computer programming and it says right here some key details of how this actually works it says like open AI Google is trying to approximate human reasoning using a technique known as Chain of Thought prompting according to two of the people in this technique which Google pioneered the software pauses for a matter of seconds before responding to a written prompt while behind the scenes it's invisible to the user it considers a number of related prompts then summarizes what appears to be the best response now Google have declined to comment on this effort but we do know that Google are actively working on this one of the things I recently covered on my channel was of course the research by Google you can see right here that this statement rings ever true you can see someone says that technically it's always been the case that Google's capabilities were topnotch they were just more conservative in Rolling things out it's a marathon and it's anybody's race to win of course remember in July Google showcased Alpha proof Which special izes in math reasoning and Alpha geometry 2 an updated version of a model focused on Geometry which is what the company debuted earlier this year it actually aced four of the six programs featured in the mathematical Olympiad an annual competition in which students tackle topics such as algebra and geometry now of course Google did give us glimpse of Astra the AI assistant which can view the world and answer questions but I think those things are still coming overall I wouldn't underestimate Google too much Google are certainly working on some incredible stuff and with their new leadership with Logan Kilpatrick I think they're going to be moving a lot stronger coming in 2025 now if we're

### Sam Altman's thoughts on AGI progress [11:36]

looking at 2025 and we're looking into the future one of the things that we like to pay attention to is of course the predictions of AI CEOs there's no other AI CEO that's more notable than the very notable Sam Alman now basically in this short clip here you can see the Sam Alman the definition of AGI matters and it means that we're getting close and it's actually going to happen pretty quickly a lot of people have various different definitions of AGI some people would call GPT 4 Sparks of AGI as they did in that early paper and some people would say that it's just completely simulating reasoning as many people don't even believe that these systems are intelligent either way we can't deny that there's going to be an increase in these models capabilities given the development speed of artificial intelligence with that being said take a look at what Sam Alman had to say at the recent open Dev day on which he actually spoke about the future progress of AI models uh we'll start with an easy one Sam how close are we to AI you know we used to every time we finished a system we would to say like in what way do this not an AGI and it used to be like very easy you like make a little robotic hand that doesn't Cube or a DOTA bot it's like oh it does some things but definitely not an AI um it's obviously harder to say now so we we're trying to like stop talking about AGI as this General thing and we have this levels framework because the word AI has become so overloaded um so like real quickly we use one for chat Bots two for reasoners three for agents four for innovators five for organizations like roughly I think we clearly got to level two we believe with 01 um and it you know can do really quite impressive cognitive task it's a very similar model um it doesn't feel AGI like in a few important ways but I think if you just do the one next step of making it you know very agent like which is our level three and which I think we will be able to do in the non-distant future it will feel surprisingly capable uh still probably not something that most of you would call an AI though maybe some of you would um but it's going to feel all right this is like a significant thing and then the leap and I think we do that pretty quickly um the leap from that to something that can really increase the rate of new scientific discovery which for me is like a very important part of having an AGI I feel a little bit less certain on that but not a long time like I think all of this now is going to happen pretty quickly and if you think about what happened from last de day to this one in terms of model capabilities and like you I mean if you go look at like if you go from like 01 on a hard problem back to like four turbo that we launched 11 months ago you'll be like wow this is happening pretty fast um and I think the next year will be very steep prog next two years I think be very steep progress harder than that hard to see a lot of certainty but I would say like not very and at this point the definitions really matter and

### Sam Altman discussing levels of AI capability [14:50]

fact the fact that the definitions matter this much somehow means we're like getting of close so there we have it a short snippet on what's Sam Alman thinks about the future of AI and one of the most important things that we did see from this interview was Sam Alman talking about level three referring to AI systems that are a lot more agentic than the ones we have now essentially just meaning that they can act by themselves and that they can go off and complete a chain of actions rather than having a simple system prompt and then one single response this means the systems are going to get a lot more humanlike Imagine A system that you can talk to and then it can go ahead and do a few different things like we have with advanced voice mode and then it can come back and say look I managed to do X Y and Z how well do you think this was done I think that once this does start to happen at some level and it's probably going to be quicker than we think because 2025 is not far away and 2025 is certainly going to be the year of AI agents overall I do think that this is quite surprising even as someone who's been covering AI it slowly does feel like the p of AI is currently accelerating whilst many Outsiders may think that things are slowing down is someone that is paying attention on the day-to-day I'm not saying that to try and build hype just based on my current expectations for where I thought certain model releases were going to be I can certainly say that things have happened earlier than I would have predicted for example video this is something that has happened a lot earlier than I thought and of course a low latency voice interaction was something that I definitely thought would happen next year at least like I said and like I said at the start of the video for those of you that haven't used advanced voice yet unless you're in the EU you definitely need to make sure you're taking advantage of this in this short clip samman actually talks about how he somewhat even got tricked by the advanced voice mode whilst even knowing it wasn't AI this is a common mistake of

### Sam Altman's experience with advanced voice mode [16:43]

humans and I say humans like I'm some kind of alien but we tend to anthromorph asize AI in the sense that we give it these human characteristics even though it isn't human at all one for me it was like the first time that I felt like I sort of got like really tricked by an AI in that when I was playing with the first beta of it I couldn't like I couldn't stop myself I mean I kind of like I still say like please toat gbt um but in voice mode I like couldn't not kind of use the normal myds I was like so convinced like ah it might be a real like you know um and obviously it's just like happing some circuit in my brain but I really felt it voice mode um and I sort of still do uh the I think this is a more this is an example of like a more General thing that we're going to start facing which is as these systems become more and more capable and as we try to make them as natural as possible to interact with uh they're going to like hit parts of our neural circuitry that would like evolve to deal with other people and you know there's like a bunch of clear lines about things we don't want to do like we don't like there's a whole bunch of like weird personality growth hacking like I think vaguely socially manipulative stuff we could do but then there's these like other things that are just not nearly as clear-cut like you want the voice mode to feel as natural as possible but then you get across the uncanny valley and it like at least the me trigger something uh and you know me saying like please and thank you to no problem probably would to do you never know now if we're talking about future predictions there's one future prediction by a CEO that has recently come back into the spotlight after a strange couple of years and I'm referring to Mark Zuckerberg his Redemption Arc is one that definitely needs to be studied and his prediction is about the future of devices I think

### Mark Zuckerberg's prediction [18:50]

the entire scene is completely changing with regards to what we view as Hardware if you haven't used meta's new Rayband glasses they are certainly a lot better than I initially thought and I genuinely cannot imagine what they're going to be like once we get AI integration Mark Zuckerberg actually talks about how smart glasses may replace phones by 2030 he says that smart glasses are going to become the next major Computing platform they will gradually replace phones by 2030 much like mobile devices surpass computers without fully replacing them I think that this is probably going to be true as technology manages to get more compact as we manage to get a lot more AI on device I do think that as long as those glasses manage to stay fashionable in the sense that they're not overly bulky and hard to wear they're going to increasingly become a part of our future if you've ever used the glasses you'll know that they're not invasive whatsoever and most people can't tell that what you're wearing is actually a piece of technology I do think however it probably won't be through meta's rang glasses one of the areas that I think is going to shift the entire Global consciousness is when Apple does glasses if you aren't familiar with Apple's brand they have a really strong brand image which means that when they usually do something provided that it's priced affordably the masses tend to flock to it and I do think that if you're in this early stage it's worthwhile experimenting with something that is probably just as good I think glasses are going to be the next major Computing platform but each new platform doesn't tend to just replace the old one right so I guess the version of this that I think about is like you probably have this experience often where you're sitting at your desk and you have your computer there yet you still pull out your phone to do things that's true yes okay so at some point in the last 10 years mobile really became the primary Computing platform we didn't get rid of our computers it's just that even when you have it you still do more things on your phone so what I think is going to happen with glasses is we're going to get to this point probably sometime in the 2030s where you have your phone with you it's but it's going to St in your pocket more because you're just going to be doing more and more things on your glasses that maybe today you would do on your phone you'll reach a point where you know just like with your computer there are probably some things that could be done in a richer way or better in some way on your phone but you're just going to the glasses will be your main Computing platform and that will be kind of your default go-to thing and then maybe over time you get to this point where people just don't bring their phone with them everywhere but I think that's really far down the line now for those of you that would like to see a product that wasn't received well for those of you that are somewhat skeptical of the f feure of AI and this combination of Hardware I can show you one recent device that necessarily wasn't for consumers but nonetheless it was still received quite badly considering the design functionality one

### Discussion of Snapchat SAI/AR [21:40]

of the things you have to understand that humans are social creatures if we all don't agree that something looks cool enough we're probably not going to use it and it's quite hard to adopt this technology without looking strange at the first a recent product that kind of suffered from this issue was of course the Snapchat spectacle now these are some AR glasses that are actually pretty well in terms of their design and their capabilities but the only problem is that they do look a little bit strange they're quite bulky and don't offer the minimalist aesthetic that meta has to offer I'm not entirely sure why this is I do know that engineering glasses that actually do perform well with any sort of Hardware is something that developers are going to struggle with considering all the engineering Feats that they have to achieve it is something that is of course to be respected but I do know that once again humans are social creatures and wearing some bulky glasses on your head is going to be something that is quite difficult if you're trying to use it in a day-to-day basis for some companies until these glasses do manage to get smaller combining them with the technology isn't going to help as much than making them more aesthetic now for those of you that have thought that Hardware isn't the future you have to understand that recently Joan I Converses involvement in an AI Hardware project with open AI the former Apple Chief design officer Joanie I has confirmed his involvement with Sam Alman this confirmation appeared in a profile of the designer by the New York Times putting to rest the speculation that began nearly a year ago I think this is

### Mention of Jony Ive's involvement with OpenAI hardware project [23:06]

a rather ambitious project as we've already SE as I was saying we've all seen how these projects can fail if you aren't familiar the recent company the Humane AI pin was one company that didn't actually succeed on their project I do think that making a hardware AI device is going to be a lot harder than most people do think one of the things I would say though is don't try and reinvent the wheel I don't really see the need for an AI pin it would make much more sense to have like a speaker at home or just have glasses like meta does considering the fact that this company charged $700 and you'd have to pay $24 a month I think this product was just too early on considering the rapid development of AI now one of the things

### Introduction of Pika 1.5 and its physics manipulation capabilities [23:44]

that did surprise me here was pabs they actually announced P 1. 5 this was one of the I wouldn't say craziest announcements but it is one that is rather underrated this new software is a lot better than people think considering the fact that they train this model in a way that allows you to manipulate physics in an accurate way later on in the video what you're about to see is that they offer ways for you to transform certain objects like melting the object which looks really cool and really nice or inflating the object which also accurately represents some kind of weird kind of physics that would happen if these objects were real you can see squishing it crushing it blowing it up is just a new kind of area that I've explored already in my community called gfx generative video effects or gfx for short this kind of thing is remarkable considering the fact that it doesn't look like the kind of mistakes that we usually see with generative AI video I think this is going to Mark the first time that we get wildd scale adoption of generative AI technology in certain workflows when trying to do visual effects you can see that inflating squishing and melting are things that traditionally would take a lot of time with normal methods but with gfx it can be done within a few seconds after having your prompt do exactly what you need it to even if you're not in the creative industry if there is one thing that you need to do it is of course prepare for the future and in preparing for the future Sam Elman gives his opinion and his advice on ways that you can actually prepare for the future of work most obvious two things to say number one you've got to learn to use the tools when I was in school you know computer programming was already popular but nothing like it is popular

### Sam altmans advice for the future of work [25:29]

today and people would say the same things many of the same things they say about AI then they would say about programming which is like oh man this is going to change things so much and make like a lot of the jobs of today go away to me the most obvious thing was well if that's the case I better like get really good at using this tool and if that's the thing that's going to like have this big impact and I think the story on computer programming has been huge that positive it has made some classes of jobs go away it's made way more new things happen and also given us the ability to do a lot of new things but it was like in retrospect that was a really good decision uh that I should like get very comfortable with that tool and I think the same thing is true now you want to get really comfortable with these new tools because we don't know exactly what the jobs of the future are going to look like I'm confident that there will be lots of jobs they and also that many of them will look somewhat different than the jobs of today but we never seem to run out of stuff to do thing number two is which I touched on a little bit I think we can say some deeply true statements about what humans seem to care about and that figuring out how to create those things fig out how to build the experiences uh products and services that people the future are going to want figure out how to be useful to people that's going to that seems like a really valuable thing more than any specific set of knowledge