What is the end game with AI ?

[Wilhelm44]

46 minutes ago, AION said:

Nah, look at all the competition between AI right now.. Shit is about to get exciting. AI helped me so much. I can't wait for the day that those stupid arrogant therapists lose their jobs to AI.

Yeah, of course not right away. ps there will always be a need for therapists that have soul.

[LastThursday]

One thing's for sure, AI will be everywhere and in everything (as if it isn't already).

In the next ten years:

It will be normal to have an AI friend or AI partner. Everyone will still have to go into the office, but we'll be talking (literally) to AI all day getting it to do things for your employers. I won't be out of my coding job, but I will be expected to be 10x more productive than I am today. The big players will still be there, but there will be a multitude of smaller players and ways to run very powerful AI for yourself "off grid". China will be big in AI, probably bigger than America. AI will get a lot cheaper to run, and per unit of compute will consume orders of magnitude less power. AI hardware compute power will increase exponentially and may well use quantum or light chips. We'll have embodied AI in very realistic humanoid robots, but they'll be the preserve of the rich and research institutions. AI will be used extensively in medicine and especially for designing very targeted drugs and doing it cheaply - possibly personalised drugs. There still won't be flying cars.

[Wilhelm44]

@LastThursday I am curious, what makes you so certain that AI wont be able to do your coding job ?

[LastThursday]

@Wilhelm44 It may eventually, but not in the next ten years. Three things I can think of:

Domain Knowledge

There's always two states that a coding project is in: starting from scratch or working with an existing code base. In either case the AI needs some input to chew on. If there's no codebase, then that input has to come either from requirements specifications paperwork or from humans telling it exactly what they want. In general both those sources are patchy and ambiguous, and so there will be a constant iteration loop with the AI to get the right outcome, there will always have to be a human in that loop - people who have domain knowledge in their heads.

If there's an existing codebase it's better because there's a large amount of domain knowledge encoded in the structure of the code base. But a human is still needed to tell the AI exactly what's required when something new needs to go in. 

In short the nature of programming will change from entering lines of code to entering lines of English (or whatever natural language), but a human will always have to check and course correct the AI, this is just coding by a different paradigm. Also capitalism always strives for the most bang for your buck, so if having a human + AI is more productive than just AI alone, then it human + AI is always chosen. It's the same reason UBI will never be a thing.

Inertia and Economics

For most companies to take on AI there will be a cost in re-organising their staff and their working practices. There is also the cost of AI per unit of time, and its effectiveness as a coder. Only when the cost drops below that of employing a human AND it is at least as effective as a human coder, then will companies prefer to use AI over people. A techie will need to be employed to set up the AI in the first place and maintain it. It will take a long time for most companies to reach that point and ten years seems about right, maybe longer - even if the AI tech is in place and is nearly good enough, which will probably happen in the next three to four years.

Edited March 14 by LastThursday

[Wilhelm44]

22 hours ago, LastThursday said:

@Wilhelm44 It may eventually, but not in the next ten years. Three things I can think of:

Domain Knowledge

There's always two states that a coding project is in: starting from scratch or working with an existing code base. In either case the AI needs some input to chew on. If there's no codebase, then that input has to come either from requirements specifications paperwork or from humans telling it exactly what they want. In general both those sources are patchy and ambiguous, and so there will be a constant iteration loop with the AI to get the right outcome, there will always have to be a human in that loop - people who have domain knowledge in their heads.

If there's an existing codebase it's better because there's a large amount of domain knowledge encoded in the structure of the code base. But a human is still needed to tell the AI exactly what's required when something new needs to go in. 

In short the nature of programming will change from entering lines of code to entering lines of English (or whatever natural language), but a human will always have to check and course correct the AI, this is just coding by a different paradigm. Also capitalism always strives for the most bang for your buck, so if having a human + AI is more productive than just AI alone, then it human + AI is always chosen. It's the same reason UBI will never be a thing.

Inertia and Economics

For most companies to take on AI there will be a cost in re-organising their staff and their working practices. There is also the cost of AI per unit of time, and its effectiveness as a coder. Only when the cost drops below that of employing a human AND it is at least as effective as a human coder, then will companies prefer to use AI over people. A techie will need to be employed to set up the AI in the first place and maintain it. It will take a long time for most companies to reach that point and ten years seems about right, maybe longer - even if the AI tech is in place and is nearly good enough, which will probably happen in the next three to four years.

@LastThursday Interesting, so not that i want you to lose your job, but what exactly would AI need to be able to do, to oneday completely take over coding jobs ?

[LastThursday]

Other than the things I've mentioned? Coding is extremely varied. On the technical side there are a large number of platforms, programming languages, operating systems, database platforms, cloud platforms, API's and so on, often multiple of those in one set up. Every company I've ever worked for has had a completely different setup. So how does an AI get to know your setup without eyes, ears and a mouth? The only way is to show the AI the setup in detail: this is the file system, this is the database, this is the code base, these are our documents and so on. It would have to be able to understand that interaction in a frictionless way (like an experienced coder does). I think AI could excel here, in the technical domain.

In some sense I would be happy to hand off the drudge work to an AI, most of it is repetitive and time consuming and boring (e.g. setting up a database). Most coding is quite repetitive: set me up a form that customer can fill out and save it to a database, and auto-email the customer with an acknowledgement, I must have done that same task many times in my career on many different platforms.

Separately, the AI would have to be proactive and interact with people. Conceivably it could do this via email or some sort of chat channel. It would have to know that John knows about the front end, Sarah knows about the back end, Jane is the boss and what she says goes, Fred is the stroppy impatient customer etc. A large part of coding is dealing with people, understanding what they want, and knowing who to talk to. Current AI is simply not up to the human aspect of communication, it's good, but it's not there yet.

As it stands AI is kind of passive, and unnuanced. It doesn't ask questions when there are gaps in its knowledge, it doesn't have a long memory, it's bad at explaining itself, it's a "yes man" and doesn't push back, it doesn't innovate, it doesn't see the bigger picture, it lies. It would have to improve a lot in those areas before it can take my job.

[Wilhelm44]

Cool, thanks

[integral]

I'm a software developer and I'm looking forward to be able to build software in a very free way where my ideas could directly Translate into features and so I don't need to code anymore.

I want my imagination to come directly to life.

[Wilhelm44]

10 hours ago, integral said:

I'm a software developer and I'm looking forward to be able to build software in a very free way where my ideas could directly Translate into features and so I don't need to code anymore.

I want my imagination to come directly to life.

"I want my imagination to come directly to life." 

Yeah, that does seem like a cool end game for AI application. (It counters the argument that AI will make us less creative).

[Wilhelm44]

@integral   Do you think the above mentioned would be possible, without getting an actual chip in the brain ?

[EdgeGod900]

Idk, but I will predict that it will be used for greed, similar to how the internet today is filled with so many fucking distractions that it so hard to use it for productivity. Early internet did not have so many clusterfucks of distractions.

Watch the greedy fucks of humanity ruin everything. They did it with programming languages too. Java, C++, etc has so many fucking abstractions that its actually so hard to fucking follow their code --dont get me started with how overrated object oriented programming is(i hope everyone who uses it imporperly, whoch is like 90% of programmers, get cancer).

They did it with the advancements in mathematics. Instead of teaching math deeply, using proofs and stuff, they give you formulas and explain them at a high level(with their impure analogies and ways of thinking) -- you know why!? Because thats how you develop rockets and military shit faster: is by memorizing these formulas rather than understanding them deeply.

 

I can't wait for karma to fucking kill these greedy and unconscious cu***. Man and then you see their linkedin profile: severly out of shape, no chin, no sense of style, etc.. sooo God Damn punchable.

Edited March 16 by EdgeGod900

[Cireeric]

On 12.3.2025 at 8:54 PM, Wilhelm44 said:

What might the world be like 5 years from now ? What might it be like 20 years from now ? Are we evolving towards greater harmony or more madness ?

That is highly dependent on how we will develop AI right now and in the coming years. If the core driver of AI development continues to be rivalrous gametheory it will probably be quite dystopian.

[Eskilon]

With current collective consciousness, the future seems more Cyberpunk than Solarpunk

[Norbert Somogyi]

On 2025. 03. 16. at 1:45 PM, Eskilon said:

With current collective consciousness, the future seems more Cyberpunk than Solarpunk

That's what I am afraid of, as well. Cheers for mentioning Solarpunk, though!

[Leo Gura]

Cyberpunk minus all the cool technology.

[Hojo]

@LastThursday you are def out a coding job. Unless you use ai and work for yourself.

[integral]

On 3/16/2025 at 1:30 AM, Wilhelm44 said:

"I want my imagination to come directly to life." 

Yeah, that does seem like a cool end game for AI application. (It counters the argument that AI will make us less creative).

I think it will make us more creative, most of the problem is that it's very challenging to get anything done, if I want to make like a 1-minute video it takes me 10 hours minimum and that's for something easy to make.

If I want to make a video game it's going to take me years.

So the barrier between imagination to creation is going to be much smaller, and people could focus only on the Creative Vision and the quality.

The reason everything is so Bland right now is because it is very time consuming to make anything of quality, so no one does it.

Edited March 19 by integral

[Wilhelm44]

5 hours ago, integral said:

I think it will make us more creative, most of the problem is that it's very challenging to get anything done, if I want to make like a 1-minute video it takes me 10 hours minimum and that's for something easy to make.

If I want to make a video game it's going to take me years.

So the barrier between imagination to creation is going to be much smaller, and people could focus only on the Creative Vision and the quality.

The reason everything is so Bland right now is because it is very time consuming to make anything of quality, so no one does it.

I'm still curious, to get your imagination to come directly to life, would you need a chip in the brain, or is there a safer way to do the whole brain interface thing ?

[LastThursday]

21 hours ago, Hojo said:

@LastThursday you are def out a coding job. Unless you use ai and work for yourself.

Not in the next ten years. My reasons are as I've written above. The uptake by business will be too gradual for complete replacement in that time, even if AI races ahead.

[Hojo]

@LastThursday If growth is exponential it will take over all coding in 4 years imo.