taladar

@taladar@sh.itjust.works

This is a remote user, information on this page may be incomplete. View at Source ↗

⁨Comment⁩ on ⁨Scientists discover that feeding AI models 10% 4chan trash actually makes them better behaved⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

There are plenty of tasks which they solve perfectly, today.

Name a single task you would trust an LLM on solving for you that you feel confident would be correct without checking the output. Because that is my definition of perfectly and AI falls very, very far short of that.
⁨Comment⁩ on ⁨Welcome to Campus. Here’s Your ChatGPT.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
There are some numbers in this blog post wheresyoured.at/openai-is-a-systemic-risk-to-the-… (and a couple of others on the same blog) and they really don’t look like OpenAI is going to last a couple of years until profitability.
⁨Comment⁩ on ⁨Scientists discover that feeding AI models 10% 4chan trash actually makes them better behaved⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
The difficult question about AGI destroying humanity is deciding whether to be afraid of that option or to cheer it on and LLM enthusiasts are certainly among the people heavily pushing me towards the ‘cheer it on’ option.
⁨Comment⁩ on ⁨Scientists discover that feeding AI models 10% 4chan trash actually makes them better behaved⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

As a standalone thing, LLMs are awesome.

They really aren’t though and that is half the problem. Everyone pretends they are awesome when the results are unusable garbage 80% of the time which makes them unusable for 99% of practical applications.
⁨Comment⁩ on ⁨Welcome to Campus. Here’s Your ChatGPT.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
The difference between AI companies and most other tech companies is that AI companies have significant expenses that scale with the number of customers.
⁨Comment⁩ on ⁨An earnest question about the AI/LLM hate⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
It is really not a big change to the way we work unless you work in a language that has very low expressiveness like Java or Go and we have been able to generate the boilerplate in those automatically for decades.

The main problem is that it is really not at all useful or produces genuinely beneficial results and yet everyone keeps telling us they do but can not point to a single GitHub PR or similar source as an example for a good piece of code created by AI without heavy manual post-processing. It also completely ignores that reading and fixing other people’s (or worse, AI’s) code is orders of magnitude harder than writing the same code yourself.
⁨Comment⁩ on ⁨Welcome to Campus. Here’s Your ChatGPT.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

Probably not going to go belly-up, in a while

Don’t be so sure about that, the numbers look incredibly bad for them in terms of money burned per actual revenue, never mind profit. They can’t even pay for the inference alone (never mind training, staff, rent,…) from the subscriptions.
⁨Comment⁩ on ⁨What was Radiant AI, anyway?⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
In fact Daggerfall was almost nothing but quests like that.
⁨Comment⁩ on ⁨The FDA Is Approving Drugs Without Evidence They Work⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
They also don’t apply the same attitude to those random sources they use instead. That is really the biggest problem with their approach. Literally going “you can’t trust anyone any more” would be better than what they do.
⁨Comment⁩ on ⁨Google confirms more ads on your paid YouTube Premium Lite soon⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
True, but my point was that even a lot of the commercial websites that do have other products do not depend on ads, e.g. Amazon and all the other stores would still be there, every company offering a paid service would still be there, every company providing a service related to their RL goods (e.g. specs, drivers, product descriptions, lists of stores where you can buy them,…) would still be there.

Advertising does not finance a very large percentage of the useful parts of the internet. And among those advertising financed websites that are useful a lot are essentially duplicates to get a chunk of the ad revenue without doing a lot of work (e.g. almost all news websites that just republish AP, Reuters,… content).
⁨Comment⁩ on ⁨Google confirms more ads on your paid YouTube Premium Lite soon⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
The internet isn’t built on it. The commercial websites who do not have other products (as in all the websites related to RL goods, stores,…) are.
⁨Comment⁩ on ⁨Sundar Pichai is vibe coding. 'It feels so delightful to be a coder.'⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
That last one is why, have you ever tried getting a bank account in the name of a script?
⁨Comment⁩ on ⁨True Wireless Power is FINALLY here (building a TRULY wire-free desk setup)⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Are you somehow under the impression that just because it is “global” there are no transmitters and receivers and distance does not matter?
⁨Comment⁩ on ⁨AI company files for bankruptcy after being exposed as 700 Indian engineers - Dexerto⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
The Indian Turk or short IT-worker.
⁨Comment⁩ on ⁨AI company files for bankruptcy after being exposed as 700 Indian engineers - Dexerto⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Asian Intelligence?
⁨Comment⁩ on ⁨In North Korea, your phone secretly takes screenshots every 5 minutes for government surveillance⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Storing individual button presses is ridiculous because that is much too low level when the apps also have much more high level information about your activities available. It is literally more useless than data you can acquire just as easily.
⁨Comment⁩ on ⁨Ai Code Commits⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Creating issues is free to a large number of people you don’t really control, whether that is the general public or some customers who have access to your issue tracker and love AI doesn’t really matter, if anything dealing with the public is easier since you can just ban members of the public who misbehave.
⁨Comment⁩ on ⁨New fuel cell could enable electric aviation⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Pressure is also bad for aviation since pressure vessels are usually heavy.
⁨Comment⁩ on ⁨OpenAI featured chatbot is pushing extreme surgeries to “subhuman” men⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Second Life might not be as universally successful as originally envisioned but it is still going 20 years later and land indeed still sells for thousands of US$ in certain popular locations so it is actually a pretty bad example of hype that was completely baseless, it was just over-hyped, not like current hype cycles that are pretty much 100% bullshit like cryptocurrencies, the Hyperloop, self driving cars or AI replacing workers.
⁨Comment⁩ on ⁨Meta shareholders overwhelmingly rejected a proposal to explore adding Bitcoin to the company's treasury, with less than 1% voting in favor of the measure⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Workarounds for the shortcomings certainly make bribes possible with crypto but not exactly easier than with traditional currencies.
⁨Comment⁩ on ⁨Meta shareholders overwhelmingly rejected a proposal to explore adding Bitcoin to the company's treasury, with less than 1% voting in favor of the measure⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
I don’t think people who give or take bribes usually like having a permanent history of the transaction.
⁨Comment⁩ on ⁨Companies are using Ribbon AI, an AI interviewer to screen candidates.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Most of them aren’t really because it is not anonymous, just pseudonymous and once you are identified your entire transaction history is public.
⁨Comment⁩ on ⁨Engineers develop self-healing muscle for robots: Device detects injury, heals it and resets to detect future harm.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
We don’t even have laws preventing real human abuse if the victim groups have no political lobby in a lot of countries.
⁨Comment⁩ on ⁨Java at 30: How a language designed for a failed gadget became a global powerhouse⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Python isn’t really strict enough to be a good learning language and Java has too much accidental complexity that literally matters in no other language.
⁨Comment⁩ on ⁨Companies are using Ribbon AI, an AI interviewer to screen candidates.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
Also consider that the same idiot decision makers have been happily applying Factory-management methods to knowledge workers for decades without noticing how badly that works.
⁨Comment⁩ on ⁨Companies are using Ribbon AI, an AI interviewer to screen candidates.⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
I sort of assumed that would be the case with cryptocurrencies too considering it is 100% scams but so far I am still waiting. And AI has far higher chances to convince idiot CEOs than cryptocurrencies ever had.
⁨Comment⁩ on ⁨Spread of sexual deepfake images created by generative AI growing in Japan⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
But sometimes when you leave people alone with their struggles they end up losing against those urges.

And nobody said that the production of child pornography should be legalized, though even talking about this reveals that we apparently don’t make a linguistic distinction between material that essentially requires sexual abuse of children to produce and material that requires nothing more than some art supplies and artistic skill. That is the part that I consider disgusting, that we apparently dislike it so much that we forget all about the actual harm the production of some of it does to actual children in our efforts to use euphemistic language to avoid thinking about it too much. It feels like the emotional comfort of those taking part in the public discourse about it is more important than actually solving the problem for the victims.
⁨Comment⁩ on ⁨Someone posted the Source Code of the IRS's Directfile on Github⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
You mean like replacing income tax with tariffs? /s
⁨Comment⁩ on ⁨Musk's Neuralink raises fresh cash at $9 billion valuation, Semafor reports⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:
You might enjoy this blog post someone linked in another thread earlier today wheresyoured.at/the-era-of-the-business-idiot/
⁨Comment⁩ on ⁨Spread of sexual deepfake images created by generative AI growing in Japan⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

My objective is the minimisation of harm. If someone uses artwork and that means that they will never touch an underage person then that is a good thing even though I may not like it.

I have never understood why we persecute and prosecute seemingly without taking this into account at all and treat someone with pedophile urges who never acted on them the same way as we treat someone who looked at drawn images and both of those the same was as someone who looks at actual images of real children being abused or someone who actively abuses children.

If anything we should try to offer the first two help in their attempts to never let their urges affect any real, existing children.

However a lot of the time it feels more like our society is designed to achieve the opposite in its active hostility to people who want to live their lives largely in places where they won’t encounter children.