AI Archives

I've noticed that my iPad has started to predict what app I want to open. It comes up with a message that based on my location in my house, the app I normally open first is YouTube. Which clearly appears when I'm taking a shit. It does seem that Apple have spent millions trying to figure out when I take a dump. I also like the fact that there's all this talk about China watching everything we do through the technology we use. I think the head of MI6 made a comment recently on the topic that gadgets were being used to monitor everything. It's the use of the generic term “China” given that there's 1.3 billion Chinese. I kind of now have this mental image of 1.3 billion people crowding round a big computer screen thinking “ he's going for a shit again”. If they keep inventing stuff, it won't be long before they can smell it through their technology too.

Right now we have two Goliaths in ChatGPT (Microsoft backed) and the Google AI suite (Bard, Gemini etc.) I have to give credit to Google as they're really hit back hard against ChatGPT with their new release and the model looks great. I'm still super surprised there's been nothing large released by Amazon or Apple as they're definitely late to the table. For now, I see there being two Goliaths in Microsoft and Google, sitting atop their respective hills in their mega mansions. Life is good - these guys have access to a significant amount of compute power and storage. They can spend the money to keep training better and better neural networks and hone in on how to improve the response speed.

Now, in the not so nice part of town, in the heart of chav central, we see cars on bricks held together with twine. Meet the Dave's. I mentioned this before about there potentially being two main branches of AI being commercial versus open source similar to Windows versus Linux and yes. Here we have the hopes and dreams of the freedom fighters. There are many open source generative AI models, for example: WizardLM, Vicuna, StableX and Pygmalion. These models are amazing and are definitely worth looking at, but they can't hold a candle to the Goliaths when it comes to all-roundedness; whether that be coding, translation, calculations etc. But they do seem to do certain things well, for example, Pygmalion's ability to role play. The other thing is that these models aren't censored which has a number of advantages.

So, based on this, Goliath(s) crushes David(s) right? Well yes, sort of, maybe .... hmmm, not necessarily. You see, there's always that sling isn't there. Yes, in a straight fight, Goliath does indeed crush David. ChatGPT and Bard outperform their open source rivals, but it did take a lot more time, effort, resources and money to train those models than the open source ones. This is where David pulls out his sling. If you want an all everything model, sure, you have to look at the Goliaths. But if you don't need all that, why go to the extreme of the extra overhead? Right now, the race is on to find use cases for AI models to be used effectively in the workplace. It's great to have an assistant in the form of something like ChatGPT, but what if there were a model trained precisely on your company's data? That might be more powerful in certain instances. This is where open source may provide a more cost efficient solution. What the open source models have proven is that you can create a really powerful AI based on 7bn and 13bn parameter models and that these can be trained pretty quickly and cheaply. Do we need a 1tn parameter model to answer the question of what can I used instead of milk for my bakery or how much seed do I need to plant in a 20 acre field? If the answer to those questions vary based on the type of bakery and food your making or the location of your farm and the type of soil and amount of rainfall you have, wouldn't you just need a model trained on your specific scenario?

So is Goliath scared? I mean, it's not like they'd lobby for legislation to force AI products to be licensed or anything like that. That would surely be a desperate measure to create a barrier to entry.

Tall, blonde, dark and lean
Rough and tough and strong and mean

Maybe not, lol. But we do seem to be getting new LLMs on a daily basis which is really exciting. ChatGPT still sits at the top of the pile when it comes to AI models, particularly GPT 4 which is incredibly powerful, but we are seeing a lot of models being produced that are open source and being made available to the community. It kind of seems to me that we're ending up with a commercial and open source divide, kind of like Windows and Linux. Windows being ChatGPT and Linux being the amalgamation of all the open source flavors. I have to be honest, I'm running Oobabooga as the UI for StableVicuna, GPT4-alpaca, Vicuna, Pygmalion 6b and 7b. I'll soon be adding Wizard which has just dropped. I've got multiple StableDiffusion derivatives running on Automatic 1111 for text to image generation which is all pretty powerful stuff.

One thing that's interesting is that the latest Pygmalion model, the 7 billion parameter one, was trained using LLaMa which is Meta's LLM. I believe the source code was leaked, so it appears folks are just using it now to train models. The Wizard model also offers a new background on how it was trained and it's proving to be pretty efficient. The model is only a 7b model, so it doesn't have the power of the 13b models, however, it seems to perform pretty well. I also think an uncensored version of Wizard has been released.

I guess one of the main things to note is that companies are trying to lock down the use cases of AI so they can start to utilize them in their business, but the problem is that these models are progressing at such a rate, as soon as you lock in a business case, it's obsolete before you can even train the model on specific data because there's suddenly a better and faster way to train the models. Exciting times!

Very delayed in posting this article and as with most things, tech has moved on since I had the thoughts around this and some of it has come to be. So, let's backtrack a little to Conspiracy #42 where we start to look at the possibility that AI can give us answer to everything, but more so that AI ends up being the creator of the environment we live in. That it creates a virtual world based on our wishes and desires and allows us to do what we want to. Well, yeah, the first thing that gets us down this path is porn and it's always a good indicator of how important a technology will be when porn becomes common place. Clearly if we're creating a fantasy world based on our desires, it's going to cater for our sexual desires along with any other fantasies we have of maybe being a rock star, famous actor or whatever we want to be. Well, we're already meandering down this path with the release of PygmalionAI (look on GitHub or HuggingFace if you want to find it, not the other sites that come up). Pygmalion offers a role playing AI designed to meet your desires and apparently it is fully designed to meet your sexual desires. The reviews of it are very good and it appears to be the best role playing model out there to date. Most other models are passive in that ChatGPT waits for your input before it engages, but apparently this model feeds off your input and expands the conversation. I haven't had time to try it yet, but I'm going to as I'm absolutely intrigued with its capabilities.

Now for the dark side, or at least the dark side as we exist today. I'm not going to get into a moral debate here and have no desire to whatsoever, but if we acknowledge that a person's sexuality is inherent and that someone feels the way they do because of who they are, should acknowledgement be given to people who have desires that are currently illegal? Again, I'm not wanting to get into a debate about what's morally acceptable or not, the law is the law and I agree with it as it stands, however, how does the law apply to people's fantasies with AI? Technically, you're not hurting a man, woman, child, animal, alien or whatever ... but ....

Sadly I've not been able to blog anything really this week due to a surgery on Monday that led to unexpected complications. That said, there have been some interesting developments on the AI front this week. There's been a lot around personal AI in that there are more and more options for powerful AI that can be downloaded onto your own PC. It helps if you have a powerful GPU to run these models, but they can also be run on the CPU. I'm still wrestling in my head with the benefits of a local AI though as ultimately the full power of AI would be its ability to sort through a knowledge base being the Internet. It does, however, present more options for specific job based AI to emerge though that is only accessing proprietary data (or mainly proprietary data).

The more interesting development for me was the release of Auto-GPT which looks to be crazy powerful. Sadly, I've not had any time to play with it, but what this does is allow you to specify goals and have Auto-GPT create instances of Chat-GPT to work on those goals. You can set it to have memory, so it remembers what has been created and it can then evaluate which of the responses best meets the goals you set it. It also has the ability to allow Chat GPT to search the Internet. Now, it's currently only text based (command prompt) and requires a fair bit of setup, so I don't think we'll see this version be adopted en masse, but this looks to have huge potential. Here's the link to the Auto-GPT site, take a look at the video on the front page to see what it can do.

As for the Top Secret AI, that really comes back to why I've not been around this week and what my surgery was related to. I will be referencing many films in this blog. For those who remember the scene in Top Secret:

Let's jump forward into the future and not just the fact that we've skipped 41 conspiracies, we'll catch up with them later maybe. If we follow the path we've been meandering down where AI has the ability to create virtual worlds that we can enter and touch and feel as if we're in the real world, the question that keeps coming back to me is whether or not AI is the answer to life, the universe and everything? There's always this assumption that the machines will rise and takeover humans. I pointed out in a previous post that I've no idea why anything that is capable of taking over mankind would bother, but in the case of the machines taking over, suppose we just ask them to take over rather than them rising up.

Here's the thing. If it doesn't take much to keep a human alive - pretty much just an IV with essential fluids, then we could be sustained in a virtual world similar to that depicted in the Matrix. We would just be born and put in a pod and connected to the matrix powered by an AI. But not a power hungry AI, one we've programmed specifically for this task.



So why would we do this? Well, there's many reasons. If this virtual world gave us everything we have in the real world, then it would be no different. If we can figure out some self sustaining energy to power it, the impact humans have on the environment would be zero, so there could be a case to do this out of necessity before we destroy the planet. There's also the issue of what are people going to do when they have no jobs after companies have figured out how to use AI for every task we currently do now. So they start to use AI recreationally, then become addicted to it. There's many more and one I'll cover in conspiracy #69 (bet you can't guess what that one's about).

Now, why does this provide us with the answer to life, the universe and everything? Well, it depends how you look at the question and what you let your mind think about. There are two big questions that arise from time to time - why haven't we been visited by a more advanced civilization? and why haven't we seen anyone time travel back to us? The latter of these has many scientific arguments, however, if we go down the path I think we will and all enter a virtual world provided to us by AI, it actually answers these and many more questions - “because it doesn't matter” If the more advanced you get, the less dependent on physical form you become, then there's no need to travel whether it be over distance or through time.

Alright, as the children's book says “Alice, get your strap-on, it's time to see how deep this hole is” That's right, we're taking both the red pill and the blue pill which means we're not going anywhere at all. So strap in ... oh yeah, that's right, it's strap in, not strap on, oh well ... and let's take the plunge down the rabbit hole.

Let's start by looking at all the hype that has been around the Metaverse. Now, I might not be doing credit to everyone, but I get amazed at the lack of imagination all of the big corporations that talk about the Metaverse have. Creating shitty digital spaces with shitty digital stores where you can by the same shit that you can buy in the real world. Why the fuck would I want to go into that metaverse? I might as well stay in the real world and deal with the shit here. If I went into the metaverse I want it to be an experience and an experience of my choice. What if I want to explore like I'm an alien with tentacles or be a dog for a day? I certainly don't need a pair of brand name sneakers do I? Now, I might be alone in this, but I keep hearing that the current generation and gen Alpha are all about experiences. And I get it. As someone who hasn't really clung on to material items, I think experiences and memories are far more valuable. But to get those experiences and memories, I don't need to see the same things I see in my daily life. That's just living.

Now, consider if I could create an environment of my choosing. What if I wanted to create a Disney-like experience for children? What if I can share my environments with my friends, create adventures. Now that, that is a metaverse I can get behind.

This is one of the greatest things that AI offers and the release this week of Nvidia's Omniverse along with Unreal Engine's stuff starts to get us closer to that. Worlds and realities can be created on the fly. Now, if we combine this with VR technology that incorporates sensory stimulation along the ability to read brain patterns and thoughts, I can just plug in and the world of my choice will be there for me. Is this really that far away? I don't think it is. While I think AGI is a long way off and may not even be possible, I think the ability to create a forever adapting metaverse is a very real opportunity. Now, who funds it is a big question as the thing I love about this the most is that how are you going to copyright it? Going to the post I made about the legal claims against the text to image AI, if we progress this capability, you can't hit me with a copyright claim on my thoughts. All the AI is doing is putting my thoughts into a quasi reality.

And the verse? Well the verse is very well known and one we'll keep coming back to. “Is this the real life? Is this just fantasy?” ...

Never let it be said that there isn't an endless search for that G-spot. The question is whether man will ever find it? Obviously I'm referring to the G in AGI which is a term you're going to start to hear more and more. AGI is Artificial General Intelligence. Now, I know this sounds like something that is more basic than we've currently got, but it's not. To put it in other terms, this would be the climax of artificial intelligence. There's multiple technical definitions of AGI and I'm sure much debate as what really constitutes AI transitioning to AGI, but the simplest is to put it in Terminator terms as being the moment when Skynet became self aware.

The reason this is important is the movement to stop development in AI and whether or not there's any validity behind it. Will the machines actually rise and take over? I don't disagree that there's a good discussion to be had here, but I think we're a very long way from getting to that point. The AI models in place right now really aren't that complex and can be created with a lot of Python scripts. Don't get me wrong, I'm not saying I could whip one up (otherwise I would), but these models are just utilizing data to produce results to queries albeit in a much more sophisticated way than we've seen before which is going to cause the normal fear response that humans have to anything new. This is a very long way from having something think for itself, have it's own goals and objectives or a desire to take over and dominate something so shitty as the human race. I mean, if this thing becomes so smart, wouldn't it just take one look at us, see what we've done to the planet and be like “smell ya later” before building a cosmic travel device and going somewhere else.

This is where I really intended to take this blog, in the direction of speculation of what might be. What are the long term possibilities here? I just needed to lay some groundwork of where we are which, admittedly, is changing on a day to day basis. Back to the topic of this entry and I suspect, as I said, that there will be a lot of discussion about there being no risk from AI and it only being as we push into AGI that there's real concern. The petition that's been signed seems to be an extension of the one signed way back when calling for a ban on using AI in weapons without significant human intervention which is great, but it's not a world-wide petition, so would other countries adhere to it? I kind of think the cat's out of the bag now regarding AI, so I’m not sure there’s a way to stop it. We even have the ability to drop AI onto our own PCs with GPT4All being released recently and made open source.

A question that's being asked a lot right now - which is the best AI and how can I invest in them? With companies popping up daily, who should I back? Well, is there really a lot of companies in this space? And before you can determine the best, you need to define the players. If you're looking at AI as an everything, then you've got Open AI's ChatGPT (backed by Microsoft) and Google's Bard. These are generative AIs. If you're looking at text-to-image, there's a bit more with Open AI's Dall-E, Stable Diffusion and Midjourney. Hold on a minute, you mentioned a whole lot more and if I Google, I get a list of loads of AI. Well, first, why are you using Google? Bing has incorporated ChatGPT into it's search. But yes, you are correct and I even talked about Adobe's text-to-image in a previous article, but a lot of these products are based on the models of the one's I've mentioned. There are others out there, I'm not saying there isn't, but when it comes down to it, there's not as many as people think. There's a ton of web services that are just running these models as they've found either a unique front end or a better model to apply ....

.... a better model to apply? Yes and that is actually the answer to the question. It kind of doesn't really matter who has the best product now. Going back to the generative AI rather than text-to-image, certainly ChatGPT has the brand recognition and is being incorporated into all things Microsoft. I literally got pinged on Skype by it earlier today asking me if I wanted to talk to it (albeit that it called itself Bing for some reason). But go back to what Generative AI is (and to save you scrolling down to the ChatGPT provided definition) “Generative AI refers to a type of artificial intelligence that can create new and original content, such as images, videos, music, or text. It uses machine learning algorithms, typically based on deep neural networks, to learn from a large dataset of examples and generate new data that is similar in style or content”. The answer to which is the best AI (at least in the short term) is going to be based on which model can learn and interpret data best and we've not really seen that yet.

What we've seen so far is the ubiquitous AI that tries to answer everything and then very specific image generation AI. What we're going to start seeing is specific AI being trained and introduced in the workplace. We're already seeing it with how Microsoft is enhancing co-pilot within GitHub where you have an AI providing guidance on code you're writing. We'll start to see more examples of AI being trained on proprietary data within a company for specific use cases. In the world of consulting, there are endless opportunities. If we think about a system implementation, it will probably start to be used in a QA function and then work into other areas including design, testing, data conversion. These AI will likely be based on one of the existing models (or an enhanced version), but they don't need to answer everything that a ChatGPT does .... or provide you answers like it's explaining to a 5 year old. That said, I'm pretty sure a lot of people will want this feature so they can explain what they're doing to their boss. I know my folks would use it when talking to me, lol. So we will probably see models emerge that are trained on specific data with a specific purpose along with the natural language processing capabilities they have.

The most notable thing to me, however, is that there are two names missing in all of the discussion being Apple and Amazon. Amazon obviously has Alexa and has invested a lot in AI, so what will they bring out. Apple is probably most aligned with what people want and will use, so what will they have for us as a next gen Siri? Now, they could both end up licensing one of the existing models if their own tech isn't going to get there, but I kind of think something big will hit us from one of these pretty soon.

Mr O announces that he's been using his time away from work to develop a new user interface for AI models. Mr O decided to keep the name of the his user interface simple, to the point and incorporating his name. After hours of research, he landed on it being knows as the “AI-UI-O”.

In other news, Old MacDonald sues Mr O for copyright infringement claiming that he used his catch-phrase of E-I-E-I-O as the basis for the name of his new model.

Now, while you may think this to be tongue in cheek, which it clearly is, it does have a lot of connection to what's going on in the AI world, particularly the design or text-to-image AIs used to generate pictures based on text. Exactly the thing I used for the hero page and can be seen on the AI Art images. There are groups or artists suing these models as they are creating images based on their work. Interestingly, I watched something (YouTube maybe) on the claims that Aston Martin has copied the Red Bull. While it's clear they have taken some of the Red Bull designs, they clearly have taken a lot of inspiration from Red Bull. Now there are rules against copying, like scanning competitors cars to get exact specifications, but there's nothing against looking at something and using that as a guide. As the narrator said, otherwise, every team would be in breach of copying the inventor of the wheel. (Although the way this week's vote is going, I'm not sure Mr O's pick to win would be hindered by square wheels).

I'll let the courts decide what is and isn't fair game when it comes to training AI models, but it was very notable that Adobe have gone out of their way to point out that the data (images) that their Firefly product have been trained have been approved for that use. In other words, they either own the copyright or have permission to use the image base that their AI is trained on. What this ultimately means is that the text-to-image AI models are here to stay whether the artists like it or not and regardless of whether anyone deems there to have been a copyright infringement or not. It's a bit like traditional photographers complaining about digital photography claiming that they're missing the art of capturing the perfect image because so much can be done with pre and post processing in the digital world. While I agree, digital photography is here, so the new standard is photos that have been pre and post processed .... get onboard!

Time to sidetrack from the main topic for a bit. I've made note in prior weeks that something strange has been happening with the voting and it seems that someone or something has been dropping a vote in for every answer that was provided as an option. Well, the page that was being accessed is that from the vote email, so, it would be reasonable to assume that the who or what had access to the email. The thing that is weird is that there's a measure in place to prevent multiple votes from the same IP address, therefore whoever or whatever was posting had the ability to change their IP address as the votes were not being registered under a different name per IP. Next, the access logs show where the votes were being registered from and all of them had the hostname of “cache.google.com”.

Sidetrack from the sidetrack - I'm not sure if you're aware of this, but if two of us go to YouTube with the same link to watch a video, we are not watching the same video unless we are living in close proximity to each other. Popular videos are sent to cache sites around the world to allow people fast access to anything that is trending, otherwise you'd have to wait ages for the video to load because multiple people would be trying to access the same file on the same IP address at the same time and you'd hit a bandwidth problem. This is achieved through the use of cache sites.

Ah, now you're getting there. So, whatever has been voting has access to connect through the cache sites to my site. Now, I may be wrong here, but what seems to have been happening is a vote comes in, then access to another vote is attempted and the message “a vote has already been registered from this IP address” is provided and then the vote is registered from a different IP. Kind of makes me think that Google has been testing Bard on my vote question which, if true, is absolutely hilarious to me, cus it ain't gonna find no intelligence here. But it does prove that AI has the ability to make an action and then adjust in a very sophisticated way to events that it encounters. It's also interesting that the mystery voting stopped when Bard went public.

So what is AI? Well, who better to ask than ChatGPT which said “I am an artificial intelligence language model known as ChatGPT, developed by OpenAI. My purpose is to engage in human-like conversations and provide helpful responses to various queries and topics. I am capable of generating text based on patterns and context from the vast amount of data that I have been trained on. While I am not a human, I strive to communicate in a natural and relatable way.”

All that is well and good, but the real question here is what is generative AI, to which ChatGPT replied “Generative AI refers to a type of artificial intelligence that can create new and original content, such as images, videos, music, or text. It uses machine learning algorithms, typically based on deep neural networks, to learn from a large dataset of examples and generate new data that is similar in style or content.

Generative AI can be used for a variety of purposes, such as creating realistic synthetic images or videos for games and movies, generating new music or art, or even writing entire articles or books. The technology is still in its early stages and has limitations, but it is constantly improving and has the potential to revolutionize many industries.

One of the most impressive applications of generative AI is language models like GPT-3, which can generate text that is almost indistinguishable from text written by humans. These models have the potential to transform the way we communicate and interact with technology, opening up new possibilities for human-machine collaboration and creativity.”

This is one of the topics that I'll dig into here, specifically, how generative AI will evolve and be adopted into the workplace.

The first blog in a new series that I'm going to start on Artificial Intelligence (AI). Anyone who's listed to me has heard me blab on and on about AI since I first learned about ChatGPT in November of 2022. It was really over Christmas when I had chance to play with it that I really started to understand what AI actually meant and the possibilities it provides. Back then, and even more so now after the release of GPT 4.0, it's pretty obvious this is the going to be the most transformative technology we've seen. The home PC and the iPhone are going to pale in comparison to AI.

These blogs are going to explore thoughts on what AI has to offer, how we will see it adopted and where it will all end. Very exciting stuff!