Dynamically generated code is not something new. It's not as common as dynamically generated content, but it has been there for a good while.
What is new is AI starting doing it, since dynamically generating code and interfaces implies a totally different thought process than producing code from start to end and running it afterwards.
The first model that attempts this is Claude Sonnet 4.5 from Anthropic—through its Imagine feature which was available for a limited time for their Max subscribers—but I'm sure others will try to push this limit sooner or later.
Source
My curiosity was stirred about this after hearing it mentioned in a YT podcast, and wanted to check this out. Since I am not a Max subscriber and was too late anyway, I won't be able to play with it to see how useful it can be at this point, or if it was only a playful experiment for now.
But with models being able to focus longer and longer on specific tasks or projects (looks like Sonnet 4.5 can do it for 32 hours straight, I heard in a presentation video on Reddit which I can't find anymore), I wonder how things will look in the near future...
Would you be surprised if in a few short years we would have games that are fully generated in real time by AI and you would not go through the same maps, monsters, routines and landscape in your adventures, even if you start over the game? I'm pretty sure that before that, interactions with NPCs and quests would be handled by AI instead of following a script.
Games will become more and more immersive, followed by the metaverse, and yeah, even if people stopped talking about that for now, it's still coming.
Since the Matrix red pilled us, some people are already wondering if we live in the Matrix. I don't believe we do yet, but if we don't, looks like sooner rather than later many of us might voluntarily choose to, especially if they won't have other things to do.
It is interesting to see how some exceptional sci-fi movies seem kind of prophetic many years later. I don't think it's that. They show potential futures modeled by our vivid imagination. Some of those potential futures are then chosen by history makers to happen, or strongly influence their views about the future and what could be possible. Unfortunately, most if not all history makers are or become cynics as they grow older and more powerful, interested in increasing their control, so we shouldn't expect many utopian elements in the future. But who knows? Star Trek had a bunch of utopian elements to it, and enough of them came true or are in the process of becoming real. Maybe it's not settled what kind of a future we'll have, whether it will be mostly utopian or dystopian. AI will be a determining factor in how the future is shaped up.
What do you think? How will things turn out?
I think the era of AI overlords being the main driver of most digital experiences is soon upon us. Just recently, there's been this NEO humanoid bot that could pretty much do many day to day house work according to the ad from the company but it was later debunked by MKHD that it was overhyped and actually a human was controlling the movement of the robot from a distance. I think that train of progress with humanoid robots being controlled by humans in some shape or form without much of the public being aware what's really going on in the backend could turn disastrous in the wrong hands. I'm still holding onto a speck of hope that our almost dystopian future ahead of us will have some elements of utopia to it, as a handle of sorts to not get fully immersed into the matrix.
That would be the least of my worries, that a human could control the robot from the other end instead of the robot being autonomous. Right now it's a marketing plot and a way to get robots quicker to be more human-like to be easier accepted. In the future I don't see this happening, and control will happen (to the extent it will) through training and enforced limitations and safeguards, rather than direct interventions, most likely.
Of course, having a house robot in your home to which even a technician could have access remotely to control even as a fail safe, can also introduce elements of risk through the invasion of privacy but then coupled with direct potential actions in your home that can also be opposite to your interests or well-being. But also in your interest. Everyone will likely have to choose whether to trust or not the new tech and to what extent. Given enough time, most people will likely adopt it, if they can afford it.
Right now, we see more evidence of the society (i.e. governments, corporations) turning dystopian than adding utopian elements. Let's see what the future holds for us!
I've played around with Bolt AI free version. Which I found very easy to use. But I soon reached the daily limit.
I think that the AI tools in GitHub are particularly interesting. And I hope to try something useful there. Also, I want to take a deeper look at DeepWiki (pun absolutely intended 😉).
!BBH
Thanks, I didn't know of it. I'm pretty busy these days, but I'll try to give it a test some time in the future.
Looks like one new direction for AIs is to build alternatives to Wikipedia. I heard Elon Musk talking about it and that they might build one because Wikipedia is biased. Now this DeepWiki. The more the better.
@vikisecrets wrote about Musk's Grokipedia. But I don't know if it helps with coding.
!BBH
I can see that happen that we go full virtual ai world in 20 years, that said if our western societies don't collapse on themselves, at this point technology will go back too... Anyway I don't think 500 years ago they even had the concept of an Ai making videogames, porn, writings etc so I think we will never be able to predict what there will be in the next 500 years
Absolutely! The scope of human forecasting of the future has diminished considerably after AI came into the picture. I'm not even sure we can "see" what will be in 20 years time.
The idea of AI like Claude Sonnet 4.5 creating code in real time sounds both exciting and a bit scary. It could totally change how we experience games and virtual worlds. I guess the real question is whether we’ll be able to guide this tech for good or let it take control.
Hard to say! Anthropic says in their article linked in the post that their Sonnet 4.5 model is the most aligned of all. The question is if misalignment is good at certain times, if alignment implies a negative impact on humanity or even at group or individual level, for example. But then, who decides that? The model itself? An overseer of the overseers? None of these options are perfect.
I think it's interesting to see how AI is progressing and how well it can code. I still think that it can get a good baseline, but I doubt it will be that precise and it will require human input for a while though.
Yep, they still need humans for oversight, guidance and tweaking of the code they produce. It may be that in the future the execution will be performed by one model and oversight and quality control will be assured by a different model.
#hive #posh