Hacker News Reader

Comment by neom | Hacker News Reader

For me personally privacy and security is less of a concern vs Ads and manipulation. I'm very very concerned that marketers will be able to insert very compelling and very nuanced advertising into our experiences with it, such that even if you're trained in the dark art of marketing, you will fall for it. We're already seeing ads in perplexity, what happens when your reply to your question about a dirty entry you wrote from your free diary app AI is: "good point Johh, and I'm sorry you're going through that, it's been proven that tea can help with that however, it's calming and soothing, Lipton is great".

I work in go to market, some people (my mum) might even call me "one of those evil manipulative marketers" - yet here I am- inclined to lobby for strong legal standards around this stuff, it should not be the wild west, that would be really bad. I'm extremely concerned about this.

View on HN

I am starting to wonder if the AI market will diverge into two separate implementations where one is the open source (open models?) that people can use and deploy will be done for people that are more technical and most people will use the easiest solutions.

I have used some of the AI techniques in macOS and while they are helpful I probably won't use the ones that call out to ChatGPT. But, I have found that using o1 preview it helps me with my side projects. On the other hand I have started using llama 3.3 and it is approaching usefulness and maybe an open model with CoT and reinforcement learning or even a new version of Qwen 2.5 I could see replacing.

The only future where this happens is if little, non- infrastructure / specialized hardware models maintain functional parity with "throw more hardware at the problem" big models.

That seems... unlikely.

And then we'll see the same evolution that built Google.

Someone with the best algorithm leverages that into a revenue stream, then uses that revenue to buy hardware, then monetizes its userbase via advertisement, then uses that additional revenue to buy market share.

I'm sure it'll make someone a ton of money, but it's not a future I'd like to live in (again).

Diversity is strength.

> On the other hand I have started using llama 3.3 and it is approaching usefulnes

cute phrasing! so what you're saying is that you're using it, but it's not useful. You can imagine it being useful, and it produces output similar to useful output, but it isn't useful.

That perfectly matches my experience with language models, going back to when I was running GPT2 locally.

Has there been any evidence that the improvement is not asymptotic? For ten years language models have been approaching usefulness.

When do we get to usefulness?

That's not what they said, and you're being purposely obtuse.

They were saying that Llama 3.3, an open model, approaches usefulness, in comparison to o1, a closed model, which is already useful to them.

As for usefulness in general, it's already here. It's not perfect, it can make mistakes, but saying they're not useful at all - especially for software engineers - is foolish. See https://news.ycombinator.com/item?id=42602940.

I think LLAMA 4 will get to usefulness ( I think I should say I have evaluated it). Long term training seems to be the best solution to solve the issues I have been facing and also more polish using CoT with reinforcement learning. I don't want to be paying for o1 since I have a homelab that could easily use up to 70B for LLAMA 3.3 or more if I can figure out how to run larger versions of LLAMA but llama can parallelize it to use all of my GPUs and inference would be faster so I would settle for that.

I can't see the (intended) endgame of AI as anything other than (1) replacing vast amounts of human labor and/or (2) inventing a new form of advertising that's conversational, obfuscated, undisclosed. Something dramatic has to justify the crazy level of investment.

Sure, most technology replaces human labor. The printing press put thousands of book copiers out of work. The word processor destroyed the entire job description of “typist”.

The promise of AI is that it will free human labor to focus on higher value areas rather than copying books or typing up the boss’ scribbled notes.

Automation is a shitty deal for the workforce because in every case those who are not made redundant are expected to create more work product for the same wage.

The introduction of Photoshop generative fill, Office 365 Co-pilot, etc means that my team will only hire one marketing grad instead of the usual three-four this year, and the one will be responsible for the same amount of work product as last year if they don’t die from the panic attacks.

In February, the excess profits go to the owners by way of a stock buyback and the 3-4 designers we didn’t hire are free to work at a coffee shop.

I’ve not experienced a single case where automation has enabled the workforce to have more leisure time, or more technically, where efficiency gains have accrued to labor rather than capital.

it’s a greed problem. No automation will fix that.

In competitive markets, the services should get cheaper, in the end, the consumer profits from AI-increased productivity.

I really think there is no way out of it. Things will continue to get worse before they get better. Eventually there will be so many unemployed working age people that violence will be the only logical course.

I could see the endgame of AI as surveillance and manipulation of information to achieve political goals. In the old days, expensive intelligence agencies and security services had to work hard to figure out targets' affiliations and beliefs, to propagandize, and to subvert opponents (or one's own society). Social media then persuaded users that it was a good idea to list their family, friends, and associates in a giant database along with their political beliefs, hopes and fears, and pictures of their lunch, and its algorithms surfaced whatever the makers wanted (cf. TikTok). Now, AI is poised not only to comb through what e-mails and files hadn't been available before, but also to surface or even generate information, giving it a prime spot to censor or manipulate what we see.

> I'm very very concerned that marketers will be able to insert very compelling and very nuanced advertising into our experiences with it...

But this isn’t new—it’s been happening for a long time without AI, and it’s scarier than we often realize. With just a few data points about your life, marketers can accurately identify your traits, even if you think you belong to a very specific niche or tribe.

What’s worse is the deceitful and unethical practices behind this: many companies have been sharing your data to build detailed profiles of you, often without your knowledge or consent.

> For me personally privacy and security is less of a concern vs Ads and manipulation.

Those two don’t add up. You expose yourself to manipulation and control when you don’t have adequate privacy. If privacy is less of a concern to you, then manipulation equally so too.

For me it's slightly nuanced. Personally, I'm less concerned about yet another attack surface for my personal information from a 3rd party I didn't expect, it's probably all out there anyway. If I gave all my data to an AI company and some hacks in and steals it, for me anyway, I'm less concerned (I realize very many people find that strange/insane). I am considerably more concerned about being psychologically manipulated by an AI tool. I would say however, to your point, it would be shit if someone hacked in, changed or inserted data, and that was use to manipulate the AI, however my sense is that might be difficult/unlikely?

Scenario: You get bloodwork done. You upload the report to an AI app and ask it for its interpretation. It says you have high cholesterol, low testosterone, and some elevated liver values.

The next day you start seeing ads on instagram for GLP-1’s, TRT, cholesterol medication, and supplements to “improve liver function”.

It may not be that the AI tool is directly manipulating you. It could simply be that your data is used or incorporated into a new AI personalization model where the manipulation happens outside of the AI tool you originally gave your data to.

Just to make it very clear: I am less concerned about my personal information being stolen, hacked, etc. I accept that stuff happens.

I tech people how to do that all day every day without AI, we have the tools to do what you described already, and we in fact indeed do what you said already.

Yes this will most likely happen. I wrote my PhD thesis exactly around this topic (~8 years ago). It was fascinating to see in a lab setting how I could manipulate humans with voice agents (robots).

> I wrote my PhD thesis exactly around this topic (~8 years ago).

For anyone interested, from checking the HN profile the thesis seems to be “The Power of Robot Groups with a Focus on Persuasion and Linguistic Cues”.

https://ir.canterbury.ac.nz/bitstreams/cac255b7-43fe-4e08-94...

Why would you do that? They could have shared a link if they wanted, they didn't.

Their HN profile is full of personal information, including name, email, location, resume, and saying they’re open for work. All I did was search for “<name> thesis” and found it without any significant effort. Why would they not want people to read their public research while talking about it with interest and making it easy to find?

It seemed more likely to me they simply didn’t want to be perceived as tooting their own horn. Turns out that was exactly the case.

https://news.ycombinator.com/item?id=42603304

You're right, my bad, sorry for the noise

Thanks for sharing my thesis. I'm never sure if I should include links or not. It kind of feels too pushy.

a cloud based AI having access to a document like a W2 pdf on my computer worries me. It has details like social security number, how much you earned, etc.

Imagine trying to perform a “right to forget” request on data that was trained into the AI?

Imagine having to comply with such request or elsr

https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-re...

It's literally a principal-agent problem (https://en.wikipedia.org/wiki/Principal%E2%80%93agent_proble...). If the evolution of Google's results page is any indication, it's a matter of "when" (not "if") the usefulness gets sold away to the highest bidder. Having billions of people rely on your system is too tempting a thing to not sell away.

"Parasitic middleman" is Silicon Valley's business model.

Reply on news.ycombinator.com