pms

This submission was about an algorithm of X, which clearly is on topic, particularly nowadays when millions of users migrate to Bluesky.

The study doesn't offer any conclusion, but it analyses some data and offers a potential interpretation. Other interpretations can be provided and discussed. The post was heading for the front page. Personally, I was very interested in the discussion.

Was any HN guideline broken?

Someone flagged this HN post without specifying which HN guideline was broken, and the post is relevant and was going for the front page, so I'm bringing it to attention.

I think the post is taken down because some people don't like it, but taking it down for this reason goes against HN guidelines. Check out this comment thread:

https://news.ycombinator.com/item?id=42160848

Betteridge's law sounds intriguing, but according to studies it doesn't hold: https://en.wikipedia.org/wiki/Betteridge%27s_law_of_headline...

At least two days in a row now, including today: https://bsky.app/profile/bsky.app/post/3lb3qyu64bs2z

I thought the conversation was interesting. This post is clearly related to the interests of HN and it was going for the front page. Which specific guidelines did it break, if any?

This could be the result of the news feed change at X and Musk's endorsement of Trump...

The open-sourcing of the X's feed algorithm is a typical Musk's smokescreen.

First, while the code was open sourced, the configs and data that determines its behavior wasn't. Any researcher that has some knowledge of the matter will tell you this.

Second, that code was last updated in July 2023 [1]. News feed algorithms are updated regularly, especially around the time of US elections [2].

[1] https://github.com/twitter/the-algorithm

[2] https://www.science.org/doi/10.1126/science.adt2983

That's the kind of products Palantir develops, but is it confirmed that they developed Lavender and "Where's Daddy"?

Are we supposed to believe in the usefulness of AI as if it was a matter of belief? The one and only Matrix to come?

Great points!

The only thing I'd add: I don't think the responsibility for lack of automation is solely on these other industries. To develop this kind of automation, they need funds and IT experts, but (i) they don't have funds, especially in the US, since they aren't as well funded as IT industry, (ii) for the IT industry this kind of automation is boring, they prefer working on AI.

In my view, the overall issue is that capitalism is prone to herding and hype, and resulting suboptimal collective decision-making.

You're comparing a satirical video from 2015 with a 2024 marketing video... Not exactly a fair comparison...

There is a difference you neglected: nobody optimizes our society to eat less vegetables, or read less books, or to disregard new year's resolutions, so the responsibility is on us, whereas social media platforms do optimize for engagement, which implies showing us sensational and shallow content that grabs our attention, without giving us enough ways to control it, so responsibility is divided...

Maybe you and me are not like that, but most iPhone users are like that and that's what Apple wanted from the start (to artificially discriminate for the good of their business)?

It sounds to me the exact opposite.

Apple's bad-faith compliance will backfire. Enough of this.

I bet OpenAI and MS do the same, but people have a positive perception of them due to the massive chatGPT hype wave.

The board had the right non-profit long-term focus, but it didn't have the willpower to communicate and realize their vision. Ilya is a wonderful scientist, but not a great leader.

I think they didn't anticipate sucha a large backlash, especially from investors. They felt the backlash threatens OpenAI, both its non-profit and for-profit arms, so they reverted their decision, which in my opinion was a mistake, but time will tell.

Until you realize that it doesn't work well with less popular videos (any items really), because "Large Language Models Struggle to Learn Long-Tail Knowledge" [1].

[1] https://proceedings.mlr.press/v202/kandpal23a.html

I don't think it's that crazy if you trust the government to use these capabilities in the intended ways, that is to catch serious crimes. The US is a very specific country, because here people have very low trust in government and public institutions, which is why public good endeavors are less developed here, while corporations do whatever they want, as long as it seems legal. Many people would argue that's a bad thing for public good...

Are you advocating for normalization of mistakes and bad driving practices? If self-driving corresponds to bad drivers, then is it useful?

If self-driving cars are really safer than human drivers, then tech companies need to prove it by releasing more data than necessary, rather than less. Why won't they release sufficient data instead of forcing the city to gather their own data? This doesn't smell well...

They are required to report every accident, but not every mile these cars drive and in what conditions, so this data is unfortunately insufficient to make valid comparisons with human drivers...

Really, the problem is the transparency of tech companies. If self-driving cars are really safer, then tech companies need to prove it by releasing more data than necessary, rather than less!

Sounds like an extremely arrogant and biased comment showing the blind righteousness of tech and its followers. How else can tech build trust with the public than via transparency? Think about this for a moment.

I didn't make that assumption. I'm just saying that overpromising and underdelivering, due to a lack of a broad understanding of AI research, can backfire, create distrust, and in the end slow science down, e.g., lead to another AI winter...

In AI research, the acronym XAI stands for eXplainable AI [1]. To understand anything about the universe, AI will need to (i) make accurate predictions about some parts of the universe and (ii) explain the laws and the models it used to make these predictions, just like natural sciences do [2]. I believe eXplainable AI can get us there one day. However, I'm surprised that, despite it's name, xAI doesn't mention eXplainable AI and the team doesn't seem to have experience in eXplainable AI, which makes me think it's just another deep-learning AI venture.

[1] https://en.wikipedia.org/wiki/Explainable_artificial_intelli...

[2] https://en.wikipedia.org/wiki/Scientific_method

When I was first reading about Musk, 12 years ago, I was extremely impressed, to the point of being ready to invest in Tesla and SolarCity. However, when Elon entered the AI space about 7 years ago it became clear to me that: (i) he is entering a field that he doesn't know, while talking about it as if he knew it, (ii) marketing and hype matter to him more than science.

Formal verification research is conducted since 30+ years and LLMs aren't really good at math and logic, so based on what grounds can we expect a breakthrough in machine guided proofs anytime soon?

So how will we know that it did?

Not just r/science. The hard part is that this logic applies also to "word/science" and to "world/society", and that this issue of belief confirmation will deepen until we learn to sufficiently value public good (knowledge discovery is a part of it) in our capitalist societies. Not easy... Unfortunately, it may take another World War for our societies to value public good more than we do currently. Call me a pessimist or give me reasons to think differently.

You're fooling yourself to think that Tesla is the solution. First, most of the fatalities are in cars that cost much much less than Tesla. Their owners will never afford a Tesla. Second, the only realistic way to prevent some of these deaths is to empower public transport, the way it's done in Europe. Third, although that's the least important point, Tesla statistics of miles driven between crashes are biased, because they turn off autopilot before crashes [1].

Please be careful not to conflate the goals of capitalism with the goals of technology and science.

[1] PDF https://static.nhtsa.gov/odi/inv/2022/INOA-EA22002-3184.PDF