Hacker News Reader

Comment by jdkee | Hacker News Reader

Someone said, "The AI you use today is the worst AI that you will ever use."

Someone could have said "The Google you use today is the worst Google you will ever use" 15 years ago and it may have sounded wise at the time.

That's a good point. But on the other hand Google search usefulness didn't have "scaling laws" / lines on capability graphs going up and up...

Maybe not 15 years ago, but they did early on. And then at some point, the lines kind of leveled off.

I can't tell any difference between Claude 3.5, 3.7 and 4 for coding.

So today is the same AI I used last year. And based on current trajectory same I will use next year.

I can tell the difference between those versions of Claude quite easily. Not 10x better each version, but each is more capable and the errors are fewer.

But errors nonetheless.

You won’t find perfection anywhere

The imperfections of people are less imperfect than the machines.

errors are all over the place no matter what. the question is how predictable they are, and if they can be spotted as they show up.

the best bugs are the ones that arent found for 5 years

There is certainly a difference, however Anthropic did really well with 3.5 - far, far better than any other provider could do, so the steps from there been more incremental while other providers have been playing catch up (for example Google's Gemini 2.5 Pro is really their first model that's actually useful for coding in any way).

If there is one cliche that I’d like to outlaw, it is this one. It’s a vacuous tautology and yet I hear so many people parrot it all the time.

There are many reasons to believe LLMs in particular are not going anywhere fast.

We need major breakthroughs now, and “chain of thought” is not one.

That doesn't take into account how it will be used against its customers once they're dependent on it. As any other service inevitably will be.

It's like if we'd said the Youtube we used in 2015 was going to be the worst Youtube we'd ever use.

Is the internet today better than it's ever been? I can easily imagine LLMs becoming ad-riden, propagandized, cloudlfare intermediated, enshittified crap the way most of the internet has.

I was going to come back here and say "what about transatlantic flight, which has gotten slower?" --- but your point is more relevant, more immediate, more impactful.

I do believe the LLMs we're using today are the best they're going to be - for the reasons you've highlighted.

Some superior tech might displace them, but LLMs as they are seem much more likely to get worse.

I'm encouraging people to make any important queries right now and save the results. For example "which books should i read on X" - right now you get good answers, in the future it'll be enshittified.

Someone also said the current AI is the less biased.

Future AI's will be more powerful but probably influenced to push users to spend money or have a political opinion. So they may enshitify...

Given that no models are profitable for the parent company afaik, it's only a matter of time before the money-squeezing begins

<think>I will reply with an example of bias in large language models. This comment seems unrelated to farmers in South Africa. </think>

Ultimately these machines work for the people who paid for them.

Reply on news.ycombinator.com