Ask HN: Are You Preparing for the Singularity | Hacker News Reader

Ask HN: Are You Preparing for the Singularity

9 pointsblululu20 comments

At this point it seems pretty clear that we are in the midst of the Singularity right now. Noteworthy unsolved math problems are being solved at an alarming rate. Agent swarms have executed hacks at the level of nation state actors. Depending on the domain, frontier AI models demonstrate intelligence at human levels or above. And capabilities are improving at a rapid rate at every level of the stack with no signs of slowing down. At this point it feels like we are witnessing one of the most significant events in world history. The risks and downsides are obviously catastrophic in nature. Job losses being among the potentially milder issues coming up. What are people doing to prepare for the transformations that are coming?

View on HN

Almost every interaction I have with these AI chatbots is still terrible.

Just yesterday I asked a question related to retirement planning and 401K allocations, Traditional vs Roth. I had a finical advisor tell me one thing, but I heard a contrary view from someone else that made some sense. The AI took the side of the finial advisor, assumed the standard dismissal of the contrarian, and that was that. I went on to explain the reasoning the contrarian gave, which was different from the assumptions, and injected some of my own thought on the topic. It then agreed that the contrarian was correct, I was correct, and the advice from the financial advisor only works under very controlled conditions, and with certain assumptions in place, and I'm better off doing what I'm doing, which fell more in line with the contrarian.

This did not feel like a chat with an intelligent being. It felt like a chat with a parrot. It starts by parroting the standard wisdom everyone knows, but frames it as gospel. Then, when challenged, it parrots back whatever it's told, as long as it's a decent and defensible argument.

For those who never push back on answers, they're going to have a bad time. For people who go too deep and use the AI as someone that affirms everything they say, they will also have a bad time. The answer to nearly everything in life is, "it depends". This is where personal opinions, experiences, style, etc come into play. The AI has none of this, so it can be pushed and pulled to any side of the "it depends" spectrum. Is that intelligence? Is there value in that?

As a tool, there is some utility. I don't deny that. As an all knowing technology that will take over the world, I'm not buying it.

You realize that most AIs will try to align with your world view right? For fun try the reverse and say you are worried about trusting your friend and wonder if you should accept the advice of the financial advisor.

That's my point. An intelligent person will adjust their views based on new information, but they're not just going to align with the worldview of whatever person they happen to be talking to. A person who does this is of little to no value, yet when AI does it people start making AGI claims.

Man, I totally get the foul sentiment, but I know myself pretty well, and I did a lot of research and learning prior to LLMs. Being ill, I have trouble retaining it though. An LLM, in seconds, can elaborate and remind me of thousands of things I've forgotten, and sometimes elaborate quite respectably, and often, introduce new material, sometimes, genuine perspective.

I keep saying, a good LLM is ultimately, or at least secondarily, a product of the user. Or maybe I should say the quality of output largely depends on the user.

That said, I typically do put a lot of work into establishing standards before reaching value. It happens though.

Strange that you say this. 90% of my Claude Sonnet 5 sessions are vicious arguments. Now 4.6, that could be if I allowed it, but then we are arguing about whether it's being sycophantic or not, a third of the time. At least I can explore some eccentric perspectives with 4.6, but I do have to stay sharp.

I think one would have to be pretty unimaginative and super corporate-friendly to align with sonnet 5.

> a question related to retirement planning and 401K allocations, Traditional vs Roth. I had a finical advisor tell me one thing, but I heard a contrary view from someone else that made some sense. The AI took the side of the finial advisor [...] This did not feel like a chat with an intelligent being

It is a great natural language ETL-type tool for all sorts of things.

Curate the words you use into directives; have it search and return data, research answers, present facts (or alternates), focus on specific areas, do comparisons, include an adversarial take, etc

But "talk to it" like it's human and you are likely leaking emotionality-based expression, so that's what you'll get back. Tell it some version of "but I heard a contrary view from someone else that made some sense" now it's triggered it into some weird role-play partner path where it outputs an argument over something it never understood in the first place, or goes the self-flagation route.

> For those who never push back on answers, they're going to have a bad time

I have trouble internalizing why "push back" or "vicious arguments" ever enter the picture. I get how it happens, people unwittingly contaminate their queries and context with emotionality and essentially end up arguing with themself. Thus fail to see why anyone picks a conversational expression to help with accomplishing pragmatic tasks when that "chat" expression style is seeking a friend, acquaintance, or adversary to continue a "conversation" with.

This isn’t what’s being sold. The OP is also saying things like, “frontier AI models demonstrate intelligence at human levels or above.”

If this is true, a user shouldn’t need to curate their words to maintain the illusion of intelligence from the system.

If LLMs are just a tool, with limits, just like every tool, then what you say is easy to accept, learn, and adapt to. For the last several years we’ve been told to bring natural language to it, the tools adapts to the user, and many have called it intelligent. People are accepting the first answers from an LLM as established truth, regardless of how the question was framed.

I’m using it how the people selling it are saying to use it. And when I do that, it’s not very good. If they want to adjust their messaging and scale back the hyperbolic claims, then I can adjust how I use it, my expectations for it, and maybe have a more positive view. Until that happens, I’ll judge it based on the high bar they’ve set for themselves, where it routinely falls short.

> I’m using it how the people selling it are saying to use it. And when I do that, it’s not very good. [...]

That is definitely one way to go about it. Going through life stubbornly insisting on using a thing only in the way it is marketed instead of in a way that optimizes its usefulness (or its cost/benefit) for you seems more burdensome than necessary. There is also a saying about the definition of insanity that may be relevant to the above

Also, my suggestion was for your use-case: a pragmatic data retrieval, parsing, and presentation task. For creative fiction uses giving directives would likely return a similarly bad experience

I have been getting flagged or near-flagged on many of my comments lately, so I will keep it short.

Approximately a year ago, perhaps closer to two, I posted an Ask HN[0].

You can either review the post, or take my word for it, but I was torn a new brown-eye for it, demeaned and even insulted for daring the ludicrous notion that such an absurdity was worth wondering about.

Fast forward to 2026, and one might now be insulted for wondering why it isn't plausible, if not inevitable to some extent.

I'd say a lot more, but I'm not in a fighting mood this evening. I'll just say that many of the folks here dismissing AI in the context of the "singularity", well, they tend to be intelligent, often technically informed, and with a fair number of karma tokens to play with. But I will tell you they are wrong. The only thing that will prevent, or defer what you are alluding to, would be a consensual pause or restriction in development. China has, thus far, made that a tricky proposition despite the efforts of Altman and Amodei et al, but consider that even if they manage to wangle a moratorium, development will absolutely not pause on the Pentagon side. Maybe the public side will slow down, but AI as 'defense', or as I prefer, weapon, will not pause for a nano second. And it will reach well beyond the extremes you suggest. And I am not sure there is anything one can do to prepare for that. My greatest concern is institutional capture, which we are already seeing.

0. https://news.ycombinator.com/item?id=32534550

Edit: All four frontier models have Pentagon contracts. Remember, Claude (Anthropic) assisted with operation Abduct Maduro, and works with Palantir. This is the direction things are moving. Just keep an eye on it, and read between the lines. I predicted this two years ago, and here it is. AI is not just about coding, or answering whimsical questions anymore.

Yeah, that is kind of my concern here. Scaling does not show any signs of dropping off. And it would be very hard to simply put a pause on any of this right now both internationally and domestically. So that leads me to wondering what my future looks like and also how I might make whatever future we all share one worth living in. I appreciate the optimism of folks out here that nothing is happening, but that seems quite foolish at this point. It is pretty clear that these machines are thinking and they are getting smarter very rapidly. How we respond to this is the great challenge of our time. Was hoping that someone out here might have a some insight worth sharing.

Alignment is a very big word, and naturally it will miss some things. Institutional alignment is where I think the general path becomes more clear, and from thereon, I see no option for correcting.

I see the best-case-scenario future as a red queen. I do not think we were ready for this.

If you find this interesting to think about, you would probably enjoy/be terrified by "AI 2077" https://ai-2027.com/

I'm purging my brain of all the useless knowledge I've accumulated over the years and extending my cognition as far as it can go by relying on AI, and I've gained enough weight to survive the first round of famines. But I probably would have done both with or without an impending Singularity.

> it feels like we are witnessing one of the most significant events in world history

Yes.

> Are You Preparing for the Singularity

No.

Are you actually using these agents on a day-to-day basis?

Yes. Everyone here is. But given the rapid pace of progress the conversation should deal not simply with where things stand but where things are moving.

No, not everyone here is.

No, for the same reason I'm not preparing for the second coming of Christ.

I'm quite sure that AI is going to change the world for better or worse - probably worse given that it's a hype-driven revolution being pushed by an accelerationist death cult of billionaire sociopaths with absolutely no serious regard for any of its many, many negative externalities - but "the singularity" is sci-fi religion. It isn't real.

Repent and believe for the kingdom of heaven is near.

https://tenor.com/view/bugs-bunny-no-no-bunny-bugs-gif-79095...

Reply on news.ycombinator.com