![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
Imitating desired realities
At work, I am on a project that uses large language models (LLMs), an instance of the modern AI fad that brought us ChatGPT and, relatedly, image generators and suchlike. Personally, I am not into gaming and I don't own computers that have reasonable GPUs, and I have little use for systems that can deliver me results that are rather more plausible than they are trustworthy. My preference remains for the more traditional kind of AI centered on knowledge-based reasoning, though I concede that deep learning boasts some impressive successes.
Now that I am digging into modern AI a little for my day job, it becomes more obvious to me what everybody else probably realized years ago. First, I find myself tempted to use it for inane questions like recommendations from restaurant menus for kinds of people. That would be a lot of pointless computation: with each of us using LLMs for whatever comes to mind, just as we might ask our friends what they think, it sounds about as bad as proof of work for accelerating climate catastrophe.
Secondly, in reviewing various available models it became apparent that
Now that I am digging into modern AI a little for my day job, it becomes more obvious to me what everybody else probably realized years ago. First, I find myself tempted to use it for inane questions like recommendations from restaurant menus for kinds of people. That would be a lot of pointless computation: with each of us using LLMs for whatever comes to mind, just as we might ask our friends what they think, it sounds about as bad as proof of work for accelerating climate catastrophe.
Secondly, in reviewing various available models it became apparent that
uncensoredmodels are readily available, there can be
guardsand such applied subsequently. Many models are probably trained on all manner of material from the Internet, some of it from the sewer. Perhaps one can buy oneself decent graphics hardware, download uncensored models, then privately indulge whatever interactive fantasies come to mind. This goes back to previous questions on the effect of people being able to play violent video games or watch extreme pornography. Are we approaching a world in which anybody can immerse themselves in the particular virtual depravity of their choosing and, if so, what does that mean for society? I suppose that we will find out.
no subject
I've dabbled with training small numerical models on the desktop and running them on embedded hardware; they were surprisingly good but they were little more than the "hello world" of ML. I'm not entirely sure what level of power is needed to efficiently train an LLM or image/video model but in my book, my powers can only be used for good.
I read the other day that the "major" LLMs - ChatGPT and similar - have already been trained on the entire available corpus of humanity. Everything they could get their virtual hands on, and I'm sure there would have been shady deals to get access to text that was not publicly online. I also fully expect that Github Copilot has been trained on everything that was in Github (including private repos? probably, who knows for sure) as well as everything they could scrape from Bitbucket, Gitlab et al.
no subject
I don't know where the boundary is between prompting and training. For instance, if I want my virtual world to feature items from my real world (e.g., people, places), can I achieve that by prompting a downloaded large model or do I then have to resort to training?
The "Digits" reminds me of the related question of where the bang for buck is in purchasing such hardware. Kind of like with cars where you don't want new-car premium but you don't want old-car maintenance costs but also in terms of if you go for mobile versus plug-in or whatever.
no subject
no subject
But...
this goes back to previous questions on the effect of people being able to play violent video games or watch extreme pornography. Are we approaching a world in which anybody can immerse themselves in the particular virtual depravity of their choosing and, if so, what does that mean for society? I suppose that we will find out.
Years ago I stumbled upon an article about a video game banned in the US, that hailed from Asia (can't remember where exactly - I think Japan, but I could be wrong about that) - entitled "Rape Me" - and the player got to portray a rapist pursuing a victim and raping them. It was awful and I was appalled it existed.
Knowing about it's existence and that virtual reality games are now available, albeit on the clunky side of the fence? It's just a matter of time before virtual reality isn't clunky at all and light weight, easy and affordable. (It's not there yet. But it will be soon. Since it sells.) And I can see people getting immersed in that. I'm sure they already are.
What does it mean? Well..
no subject
I am reminded of the anecdote of the actor who had to play someone important so, to help them be in character, the cast and crew generally treated them as if they really were rather important, which became a challenge for the actor after the job was done and they were back to being unimportant. One world will bleed into another somehow, the question is how.
no subject
no subject