All We Need to Expect From AI & AGI in 2024! Everything has changed...

year and in this threat later he made a shocking statement AGI is not coming anytime soon at least in 2024 but is everything really that bad for open Ai and what can we really expect to see in 2024 now first let's rewind the clock and see the whole story so on December 23rd there is been a tweet from Sam asking what would you like open AI to build slf fix in 2024 there were many replies from people and just two minutes after Sam wrote this way more requests in the first two minutes for AGI than expected I'm sort to disappoint but I do not think we can deliver that in 2024 this begs the question when and what problems could open AI be facing right now AGI or artificial general intelligence according to Wikipedia is a hypothetical type of artificial intelligence that could learn to accomplish any intellectual task that a human being can perform AGI would have the ability to reason solve problems and learn from its experiences just like human Sam Alman himself once gave an interesting definition of AGI so let's take a look AGI is basically the equivalent of a median human then they could like say do anything that you'd be happy with a remote coworker doing like just behind a computer learning how to go be a doctor very competent coder like there's a lot of stuff that a median human is capable of getting good at is not any particular Milestone but the Meta skill of learning to figure things out and that it can go decide to get good at whatever you need and then Super intelligence is when it's like smarter than all of humanity put together the equivalent of a median humans seems like Sam has huge plans for AGI but what's stopping open AI from creating AGI in 2020 4 because let me remind you that recently there have been talks about project qar at open AI that supposedly has successfully created the internal AGI yet we still don't know how open AI is doing this and which approach they've chosen in general

there are four ways theoretical ways to create AGI let me brush over them real quick one brain emulation brain emulations is all about simulating those intricate neural connections and processes of the human brain this may seem oddly familiar to you because that's the basic principle behind AI but I'm going to come back to that a bit later second approach symbolic AI is all about teaching computers to think using symbols and logic similar to how we solve problems it's like giving AI a set of tools to understand and work through complex ideas making decisions like a human would third method Subs symbolic AI basically the same thing as symbolic AI but with one notable difference it doesn't use Clear symbols or rules instead it looks at huge amounts of data to find patterns just like how we learn from experience it's good at finding Connections in big data sets that humans might miss and fourth method hybrid this means mixing symbolic and subsymbolic AI along with other techniques this blend could help AI understand complex ideas and adapt to new situations much like how our brains work we don't know which method open AI has chosen but I believe that the way Google does things could give us a hint we've recently done a deep dive into Google's Gemini model what it can do and so on but I don't believe that I properly explained to you how the Gemini could be the key to AGI but before I do let's take a short pause and talk about autods the sponsor of today's video autods is an awesome Drop Shipping tool for eBay Amazon and Walmart it simplifies adding products from places like CJ dropshipping and AliExpress and managing inventory and prices what's great about autods is its automation it handles tasks like product listing and Order fulfillment freeing you up to grow your business it also allows selling on multiple platforms increasing sales opportunities autods also includes quite a few features for finding profitable products and good suppliers in this way it helps keep your prices competitive with market trends the marketplace offers a wide range of products and a winning products feature helps identify top sellers the Tik Tok

spy feature is pretty unique it lets you see what's trending on Tik Tok so you can find cool products to sell important products is easy too whether it's one item or a bunch of ones and if you want to keep things really simple autods can handle your orders for you there's also an autods s Academy that offers courses for drop shippers of all levels overall autods is a fantastic tool for making Drop Shipping more efficient and profitable I will leave a link to autods in the description so be sure to check it out just like humans AI models are constantly learning and adapting to new information and while human intelligence is multifaceted AI models have traditionally focused on specific tasks such as image recognition or natural language processing however ever to achieve artificial general intelligence we need to develop AI models that can process and understand information from multiple sources just like humans do this is where multimodality comes into play multimodality is the ability of an AI model to process and understand information from multiple sources such as text images audio and video that's exactly what Gemini was trained to do this Tech should allow AI models to grasp the nuances of the world around them just like humans do when they combine information from their senses for example when we read a text description of a painting we can also see the image in our mind and we can hear the sounds of the artist creating the painting this ability to combine information from multiple senses is essential for human intelligence it is also crucial for AI if we want to create truly intelligent machines and I think that since Google has been researching this stuff it couldn't be a far-fetched to suggest that open AI is moving in the same direction but wait I still haven't explained how multimodality Works in practice basically Gemini and systems like it use a variety of different separately trained AI models that all work together usually under the supervision of another model that was trained specifically to govern all other models it's like hierarchy with one model being the overseer and others the employees and this is one of the most straightforward

ways to reach something like AGI the main reason being the relative Simplicity of this approach I don't mean that it is a simple task in any way but combining many AI models with separate training and algorithms to work as one technically is easier than programming one model that is capable of doing everything at once all by itself again we still don't know how open AI plans to do things they can do the same thing Google does or choose a more complicated way like stimulating the brain but okay if AGI is not coming in 2024 so what is what can we expect from Sam Alton in the same Twitter X thread Sam wrote a summary of popular requests and it's quite interesting I find the text on this screenshot really important because Sam definitely had reasons to choose these things and would have no reason to choose something he hasn't thought about so I suppose this is basically a complete list of all the things that we're getting in 2024 one AGI we've talked about that already and Sam asked for patience we'll give him that surprises to GPT 5 so much has been said about gbt 5 already but one thing is clear gbt 5 is coming in 2024 and it's a big deal in the AI world it's going to be a major upgrade from Chad gbt which uses gbt 4 expect gbt 5 to be larger more powerful and way better at creating text on any topic Sam Alman the head of opena I hence that it will have cool new features and functionalities that will make using it a great experience three better voice mode yep Chad gbt can speak but this feature has been only available in mobile devices and apparently we can expect it to come to the web app too and it's huge I can say that I'm tired of typing my prompts but it would be really nice to just say what I want I also hope that Chad gbt will sound more realistic and match tone and mood rest in peace Siri and Google Assistant fourth change higher rate limits in chbt control how much text you can generate in a certain time this limit has increased a couple times already and gbt

4 now can generate more than any other language model and these limits are going to be increased in 2024 once more allowing us to do more with Chad gbt faster five better gpts we've made a whole video about my gpts what they can do and how they can be the future of AI so be sure to watch it but long story short these are customizable Mei Chad gbt that can do a variety of small things tailored just for you text generation images web search and so on and in 2024 we can expect them to get even cooler generate better text and handle more complex tasks like reasoning summarize and translating and answering difficult questions W that's a lot of changes six better reasoning here everything is in the name nothing else to add seven control over degree of wokeness behavior this one sounds really interesting because I personally haven't really noticed any issues with wokeness of chbt but apparently chbt can be a bit too left winked so it could be great to control how woke or what kind of behavior it will have this means we would set it up to match our preferences whether we wanted to be neutral more conservative or liberal or even have a humorous or polite tone but that's just predictions Sam didn't really clarify this point eight video I feel like this one doesn't need an introduction it could be two things in my opinion either generating videos or understanding videos liking that video from Google that turned out to be fake but can you just imagine asking chb2 to analyze the video generate story boards extract b-rolls and do all sorts of things call me a dreamer but I think if chbt was capable of working with video files it would be such a huge selling point that it could probably kill the majority of plugins and video focused AI of course this could end up being something like Dolly AKA generating videos from scratch and I would still be happy because if the consistency and quality will be the same as with Dolly then there will be basically no point in pain for stock videos every month it would be so easy to just ask Chad gbt to generate a bunch of videos with all the necessary visuals

honestly I haven't been this excited about the year for quite some time nine personalization here I think everything is also pretty trivial and understandable Sam doesn't clarify anything in his tweet but we can deduce that he will give us even more capabilities to personalize Chad gbt and his responses this year we already got custom instructions and if they get an update with even more variables it could be a huge shift maybe we could see something like a personal vocabulary where deep customization on the style that Chad BT uses to write our text and so on and so forth and all those copywriters out there be aware of this you better start learning new skills like programming oh wait chbt will do that even better than it can now gosh the more I talk about it the more it seems to me like soon AI will do all the work for us and there will be no things left for people to do okay 10 better browsing apparently Chad will learn how to properly search and access information from the internet which could make it even smarter I can already see students getting excited thinking of all those research papers and home assignments that they would complete in minutes with this improved model it would be nice to see gbt 5 learning how to assess the information its quality and combine the authentic pieces ignoring all the false information because right now chbt can still struggle with information even if it's something simple it can give false responses forget to correct itself and be pretty unobservant let's just hope that 2024 will'll fix that wait what is that little message from Sam at the bottom we'll keep reading and we will deliver on as much as we can and plenty of other stuff we're excited about and not mentioned here plenty of other stuff not mentioned here Sam Sam what are you hiding from us if that's not AI then what I think we need to be reasonable here and look at what can really happen at least for now the most plausible thing to happen is the final launch of the delay GPT store if you don't remember right after the introduction of my gbts open AI promised to launch a

store where people would be able to sell their specifically trained models but it never happened and there were no comment from open AI however this feature is definitely the closest one to being completed and ready for launch chbt already has something similar to that but all the models there are free and available to all users which doesn't really look like a proper store combined that with all the previous leaks about improved gpts and the picture becomes clear many people have already created their custom gpts and with the improved capabilities they will further hone them adding more and more features finally turn the gpts into something that can be worth paying money for because I have already tested dozens of different gpts and none of them was good enough to convince me to pay even a dollar for so I expect this store when it launches to have a bunch of really powerful and smart gbcs another thing that I hope to see resolved is the limit of messages for GBC 4 currently the limit is set to 40 messages per fre hours the main reason for it of course is the lack of processing power open AI Computing Department definitely cannot keep up with the demand that people have for gbt 4 so I think one of the focuses for open AI this year will be optimizing the model further to the point of it having zero limits whatsoever just like the GPT 3. 5 of course this would need substantial financial investments but judging by all the drama that we saw a few weeks later with Sam Alman being fired and rehired with the help of Microsoft open AI definitely has the money for all the necessary upgrades I just hope this comes sooner rather than later because gp4 is really cool and I get pretty frustrated with these limits when I try to do serious work so Sam please fix that another thing that was found hidden in the Chad gbt client is called project sunshine and apparently it is a special version of Chad gbt with memory and by looking at the leaks we can see just how huge this thing could be let's just look at the video and read the notes from the developers your gbt will carry what it learns between Chads

allowing it to provide more relevant responses as you chat your gbt will become more helpful Rembrant details and PR references your GPT has been designed to follow your instructions in chats you can reset your gpt's memory or turn this feature off in settings next the video shows us the so-called temporary chats here is what the text says about it temporary chats won't appear in your history and your GPT won't remember anything you talk about we may still keep a copy for up to 30 days for safety purposes your GPT won't be aware of previous conversations it will still follow your custom instructions if they are enabled temporary chats won't be used to improve our models and I already like the sound of that right now all the chats are used to train the model and all those false instructions and bad proms that we write sometimes lead to the model being sluggish and responding incorrectly but just like I mentioned before in the settings there would be a button to reset the memory and what's even better there is a special button titled memory off I guess it's supposed to turn off the memory of your personal client but still improve the overall GPT it all sounds too good to be true but I seen no reason for it not to exist my GPT has gotten pretty tired and dull after a year of active use and sometimes this responses could be far from perfect or even okay so if project Sunshine eventually gets released it's going to be a life-changing feature for me of course we can put on our tinfoil hats and start purely guessing what the year holds for is fantasize about super intelligence the rebellion of the machines and so on but let's keep our heads cool and our minds open for new things and of course let's keep the passion for AI hot in our hearts thank you so much for watching and I'll see you in the next one peace