Love the challenge, but im sorry eliminating error checking is cheating and means its not a proper self contained interpreter, because it requires a human to pre-verify and enter only correct programs.
When human beings write we do so with a particular perspective with an intention to communicate to a particular audience. Even the driest scientific writing is partially informed by the experiences of the researchers, no matter how hard they try to remove it. Human writing has a viewpoint.
LLM generated text always lacks these elements. It's always some bland, shapeless sequence of words pulled from the ether. I'm sure with time and effort you could combat this and give LLM text more of a human feeling by giving the LLM enough context about your communication, but that would take work, which in many cases would defeat the purpose of using the LLM instead of doing the writing yourself. And if you care that much about your communication with others (and you should!) you'd probably just find using LLMs frustrating to begin with.
Given how many times Trump declared the war a victory, I'm also curious to know if they declare it a war, the families will be compensated for a war, ten wars, a hundred wars, fifty wars and fifty civilization ending operations, two wars a hundred ceasefire conversions (?) nightly matches that did not break ceasefire? How many exactly?
> We don't have any proof of that, but the approximator thing is proven.
Not in the way suggested... It's not approximating human intellect. They approximate the target function, and the target function frontier labs are trying to approximate is ultimately a super intelligence...
> Synthetic data derive from other linguistic data. Whatever intelligence is in there, it is expanded horizontally, not vertically
I'll assume we're talking purely about language models for a moment, but if you assume that everything can be represented linguistically, then in theory there is no upper-bound on what can be learnt with synthetic data.
Why not just hide/flag the story and move on with your life? No one is forcing you to comment. And apparently the rest of HN disagrees with you, it is on the frontpage after all which happens after we all vote or don't vote for it.
If I didn't knew better, it does sound like you're slightly jealous of something...
I'm sure it's also a matter of how much time you actually dedicate daily to this - and it's quite hard to maintain passion and focus for 8+h /day on the same game for 30-40 years.
"OpenAI’s primary bet here has been chain-of-thought monitoring. It is based on an appealingly scalable idea: a lot of the model’s capability comes from a verbalized reasoning process (chain-of-thought). If we scale optimization on the outcomes of that process, but do not supervise the process itself, that chain-of-thought has no direct incentive in training to hide any misaligned ideas or objectives. This does not mean the model will learn to externalize misaligned tendencies that don’t rely on using the chain-of-thought; however, it can allow us to monitor exactly the capability increase from reasoning."
Yes, and it would greatly aid alignment if the user could monitor the chain-of-thought! The open models, including very powerful ones like Kimi K3, are delivering this. The fact that OpenAI and Anthropic are not clearly indicates that a commercial consideration (avoiding distillation) takes precedence over alignment -- regardless of how much they bloviate about the latter.
Nah mate, I have put my time into the 'third sector' (charity/aid) with organisations that don't patronise and condescend. USAid is utterly vulgar, although some of it might be cultural, for example, Americans have 'Mr Beast' as their model of philanthropy, which is utterly cringe if you don't buy into American capitalism.
Besides, your comment adds precisely nothing. Cite one African by name that did not starve due to the wonders of USAid. Or even one Afghanistan person. I will wait.
Oh for God's sake, can the Europeans please make a European HN already, so these incessant stories go there?
I am genuinely happy for you, I wish all you the very best, more competition is good for everyone, and now please stop telling me about your upcoming Swabian clone of MS Paint.
It's not actually hacker news. It's not even tech news. It's European bureaucracy news.
(Rookie mistake, posted this in European daytime hours during the summer holiday months of February-November where all the Europeans are downvoting anything they don't like hearing.)
Watching the actual investigation video the article is based on, it turns out that alot of the statements in the article don't really hold up to closer scrutiny.
I couldn't find any "smoking gun" or discovery of malicious intent.
What's left in the end is the already-known fact in the tech-community: The most-common, most-vulnerable and most-equipped device to spy on you in your home is your Smart-TV.
1. It has all the compute-power and sensors it needs
2. It already does some spying for ad-profiling
3. If a hacker exploits it, it's a problem
It's insane to me how you can defend Russian imperialism while criticizing American imperialism in the same sentence. Moscow invaded Ukraine twice in the last 15 years. So yes the current Russian regime is an enemy to Ukraine, Europe and the US. Russia is a master in playing the victim and blaming its internal problems on external forces. Putins crying about NATO expansion and foreign agents manipulating the population was basically just projection because that is what Russia has been doing since the times of the Sowjet Union.
Regarding USAID: During researching topics in health care the research was often funded by USAID and these topics were definitely interesting and good and allowed smaller startups to exist. So this funding did a lot of good and created technology and research jobs in the US that are probably gone now.
Creator here. This site is a clone of outbid.lol, with one exception: you must have an EU VAT number to bid. This number is verified with VIES before payment.
How it works: Your ranking is based on the total of your payments. Bids are permanent; increasing your bid only costs you the difference. Equal bids are awarded to the oldest listing, and reaching first place costs at least €5 more than the currently top-ranked listing. That's the ranking logic; there are no other factors or rating system. The rules are available at overbid.eu/rules.
VAT management: The VIES REST API is called before a Stripe session is created. The results are cached so that top-ups are not redirected to VIES. Invalid results are cached for 24 hours to prevent typos from being blocked, and transient VIES errors are never cached. VIES is known for its sometimes unstable reliability depending on the member state; therefore, a retry system with a timeout and a "Notify me when I can try again" field are implemented. Verification is systematic. The legal name and country are taken from the VIES response and are displayed on the listing.
Configuration: a single Go executable, SQLite in WAL mode with a single-write connection, templ + htmx, Caddy as the front end, on an OVH VPS in Gravelines. The forum is served from a snapshot in memory that refreshes every 5 seconds or with each paid bid; the database is therefore not accessed with every request. Payment via Stripe with VAT self-assessment: self-assessment is applied when the VAT number is valid.
I am available to answer any questions regarding VIES integration or SQLite configuration.
> In November 2023, Meta introduced a binary ‘Consent or Pay' advertising model. (…)
> The Commission found that this model is not compliant with the DMA, as it did not give users the required specific choice to opt for a service that uses less of their personal data but is otherwise equivalent to the ‘personalised ads' service.
