I’m glad someone pointed this out about the case study/brain teaser interviews:
> “Many wealthier students have a sense of this, he says, from parents or networks, but to low-income students—save the handful who receive coaching—it’s utterly counterintuitive. “This is one of the major ways that consulting firms—and, really, investment banking firms—block low-income students,” Portela said. For all practical purposes, the case study “is in another fucking language.”
Being raised on welfare and paying for college on my own, I hated the post-grad interview cycle because it seemed so asinine. Almost 20 years later, I’m somewhat better at knowing how to answer these questions, but I can’t help but to acknowledge the cynical side of these questions in finding surreptitious ways (either intentionally or unintentionally) discriminate against low-income candidates.
I’ve been thinking a lot about a similar concept, but orthogonal application of that concept: when immediate/short-term incentives are not there, how do you reward workers in the trenches (scientists and engineers in this case) to push forward and make the best decision for science, even if it’s not the best decision for the business/entity?
Isn't that kind of the default position for these types of builds? Nobody wants to be the team that built a thing that died on day 101 when the mission was designed for 100 days. Everyone wants the science to stop not because the platform stopped working but because bean counters shut it down. Everyone wants to be the team that built Curiosity long outliving its mission duration while continuing to do science. Otherwise, the bids will go to the teams that build Voyagers or Curiosity and never come to the team that builds systems that last exactly mission duration
Eh, idk who Hinton is, but I’d cut him some slack for making both statements- I could imagine a case where “creatives” can semantically be understood as “new blue collar.” Musicians, dancers, photographers… are not blue color manufacturing employees, but they are fiscally more similar than their white collar counterparts. It’s possible he used inconsistent terms because he really means “low-wage employees who are far away from the monetary benefit creation decisions,” but that’s a mouthful
Hinton is the guy from the article. He is a big figure in AI research.
For context: he once argued AI could handle complex tasks but not drawing or music. Then when Stable Diffusion appeared, he flipped to "AI is creative." Now he's saying carpentry will be the last job to be automated, so people should learn that.
The pattern is sweeping, premature claims about what AI can or can't do that don't age well. His economic framing is similarly simplified to the point of being either trivial or misleading.
Carpentry is already partially automated. I’ve worked on cutting algorithms to minimise waste. There are a number of startups which will go from a 3D interior design to manufacturing. Think of customised Ikea.
Ehhh it sounds like he's a poster boy who rode on the success of others (LeCun, Deepmind) and says whatever the current popular opinion is until proven wrong and shows no hint of predictive capability.
Hinton published the seminal paper on backpropagation. He also invented Boltzmann machines, unsupervised learning and mixture of ecperts models. He championed machine learning for 20 years even though there was zero funding for it through the 80s and 90s. He was Yann LeCun's PhD adviser. That means Yann LeCun didn't know ass from tea kettle until Hinton introduced him to machine learning.
Know perchance a fellow by the name of Ilya Sutskever? ChatGPT ring any bells? Also a student of Hinton's. The list is very long.
Frankly, this all sounds like hero worship and the language is very cringe.
I know the backprop paper. I've read it in the early 2000s. And I remember Hinton as a co-author. Same with Boltzmann machines. Co-author. "Advisor to that great guy", "Teacher of this great guy", "Nobel price together with that guy" <- all of this leads me to the above conclusion. YMMV
just one example of the halo effect: having been instrumental in the development of an important technology doesn't magically make one an expert in the economic impact of that technology, as economy is a completely different field of study
I'm not a fanboy, far from it. I'm not affiliated with the lab or his work. I'm not even a big fan of machine learning. But Hinton's contributions to the field cannot be understated. He single-handedly kept it alive for two decades amid a massive lack of funding. Anyone who has worked in research will attest that this is an incredible feat.
People on Hacker News seems to idolize the lone genius who somehow pulled himself up from his bootstraps. That person does not exist. The truth is that great minds are made, moulded into shape. That the best people behind our AI technology emerged from his lab is no coincidence. Those trash-talking Hinton on this forum are unlikely to achieve 1/100th of what he has accomplished.
“The housecat may mock the tiger,” said the master, “but doing so will not make his purr into a roar.” [1]
Ah, someone who enjoys defending his tigers against cats. Well, enjoy your low hanging fruits, tiger. Someone "so knowledgeable" and totally not a poster boy making bold claims that don't come true at all will continue not making sense to me. But don't get me wrong, I know this is an old man shakes fist at cloud kind of situation. Who cares what doesn't make sense to me anyway? My academic circles don't, and they do fine. I continue to contribute what little I can. But my field is more SR than NN.
idk… I for one like the fact that Gmail sucks at search to be honest because I (naively?) believe they don’t profile everything in your inbox. For example, if I subscribe to a newsletter about “healthy lifestyle,” it won’t return that newsletter, but return string matches where “healthy” or “home” are relevant. If they profiled the emails for contextual awareness to know what I meant by “healthy living,” I’d be concerned.
Well, if I want to first understand the basics, such as “what do the letters OSINT mean,” I’d think the homepage (https://osintframework.com/) would tell me. But alas, it does not, and a simple chatgpt query would have told me the answer without the wasted effort.
Similar criticisms that outsiders need to do their own research to acquire foundational-level understanding before they start on the topic can be made about other popular topics on Hn that frequently use abbreviations, such as TLS, BSDs, URL and MCP, but somehow those get a pass.
Is it unfair to make such demands for the inclusion of 101-level stuff in non-programming content, or is it unfair to give IT topics a pass? Which approach fosters a community of winners and which one does the opposite? I'm confident that you can work it out.
OSINT = open source intelligence. It’s the whole of openly accessible data fragments about a person or item of interest, including their use for intelligence-gathering objectives.
For example, suppose a person shares a photo online, and your intelligence objective is to find where they are. In that case, you might use GPS coordinates in the photo metadata or a famous landmark visible in the image to achieve your goal.
On which platform? It's a mature project that has been working for years on desktops and phones, with content coverage that has expanded beyond wikipedia, e.g. stackoverflow archives. Downloadable from the nearest app store.
Volunteering “I give up if the information I want isn’t on the first page of the first website that I think of” in a thread about AI tools eroding critical thinking isn’t the indictment of the site that you linked to that you think it is.
There is a whole training section right there like you just didn’t feel like clicking on it
The OSINT framework isn’t meant to be an intro to OSINT. This is like getting mad that https://planningpokeronline.com/ doesn’t explain what Kanban is.
If anything you’ve just pointed out how over reliance on AI is weakening your ability to search for relevant information
I just tried reading the documentation and I have no idea what I’m supposed to do. I have a bunch of keyboards that won’t let me use the F keys as standard f keys (only lets me use them as multimedia keys) unless I hold the keyboards fn key. Karabiner is set to “use f keys as standard f keys” and it still doesn’t work. Anyone know how I might use kanata to use F1, F2…etc without holding the keyboards fn key?
I don’t know Kanata, but you need to find out which key codes are sent for the multimedia functions in order to remap them. This file [0] (linked from [1]) contains the key names Kanata knows, and it has entries like “VolumeUp”. You could try to remap those to the respective function key.
There are some keyboards, however, which send multimedia commands via USB “out of band” from normal key codes, and which therefore cannot be remapped that way.
Which is nuts for a simple article page like this (and almost any other use case really). Just leave the native components as they are, especially scrolling.
Firefox on macOS is also completely borked. At least with my freescrolling Logitech Master MX 2. I'll scroll and it will jump the expected amount, I'll scroll some more and it will jump about 4-5x what I expect, and then I'll scroll some more and it will go backward in the page. The site is pretty much unusable for me.
I would just kill the scrollbar styling if it’s breaking the site in common browsers. Styling the scrollbar is absolutely a nice-to-have—more important to have people able to read the content. Otherwise, what’s the point?
Isn’t Deepseek’s advantage that they didn’t actually start from scratch and that embeddings & training was already supplied? To say that Deepseek is comparable in performance to OpenAI is like saying Kirkland brands is comparable to {insert non white labeled good here}- they’re created off the same inputs. To say that Deepseek is a threat to AI supremacy is hyperbolic. As long as OpenAI innovates at the rate that it’s been innovating, their value is undeniable. Sure, Deepseek may tick after OpenAI’s tock, but the premium is in that tock.
This take discounts the implications of the fact that the efficiency and cost of Deepseek's product are orders of magnitudes better, plus the project is open source. Those facts fundamentally shift the business conditions for LLMs.
when I asked for his support around difficult prioritization discussions, his response was “I believe in the Team First mentality, and our success is not necessarily putting my team nor our customer first- it’s in putting my peers’ needs first. Do whatever it takes to make them successful.”
On one hand I could see how this can theoretically lead to good outcomes all around… on the other, is this a joke? I hope to god this mantra doesn’t catch on, because what is a boss if he is not there to support you.
> “Many wealthier students have a sense of this, he says, from parents or networks, but to low-income students—save the handful who receive coaching—it’s utterly counterintuitive. “This is one of the major ways that consulting firms—and, really, investment banking firms—block low-income students,” Portela said. For all practical purposes, the case study “is in another fucking language.”
Being raised on welfare and paying for college on my own, I hated the post-grad interview cycle because it seemed so asinine. Almost 20 years later, I’m somewhat better at knowing how to answer these questions, but I can’t help but to acknowledge the cynical side of these questions in finding surreptitious ways (either intentionally or unintentionally) discriminate against low-income candidates.