HNHacker News
TopNewBestAskShowJobs

ip26

9,764 karma · joined September 26, 2018

thebass (at) gmail
submissionscomments
ip26··on Why friction is necessary for growth
Its "product recommendation" capabilities are growing quickly, which is directly convertible to ad revenue, and we know ad revenue sustained Google for two decades.
ip26··on ChatGPT Pulse
It doesn't have to be that way of course. You could envision an LLM whose "paperclip" is coaching you to become a great "xyz". Record every minute of your day, including your conversations. Feed it to the LLM. It gives feedback on what you did wrong, refuses to be your social outlet, and demands you demonstrate learning in the next day before it rewards with more attention.

Basically, a fanatically devoted life coach that doesn't want to be your friend.

The challenge is the incentives, the market, whether such an LLM could evolve and garner reward for serving a market need.

ip26··on ChatGPT Pulse
Argue with it. Criticize it. Nitpick the questions it asked. Tell it what you just said:

you just gave me a generic summary of the top 5 articles you'd get if you googled "how to fix depression, social anxiety" or something

When you open the prompt the first time it has zero context on you. I'm not an LLM-utopist, but just like with a human therapist you need to give it more context. Even arguing with it is context.

ip26··on Analyzing the memory ordering models of the Apple M1
I’m not an expert… but it seems like it could be even simpler than program design. They note false sharing occurs due to data not being cacheline aligned. Yet when compiling for ARM, that’s not a big deal due to WO. When targeting x86, you would hope the compiler would work hard to align them! So the out of the box compiler behavior could be crucial. Are there extra flags that should be used when targeting ARM-TSO?
ip26··on Betty Crocker broke recipes by shrinking boxes
The no-frills commodity mixes often seem quite cheap so it’s possible the price was still pretty fair.
ip26··on Betty Crocker broke recipes by shrinking boxes
Some pack, some sift and scoop. You have to know what the recipe wants. Weigh is in my opinion superior, it just wasn’t widely accessible before cheap digital scales.
ip26··on Betty Crocker broke recipes by shrinking boxes
There’s a standardization element. You can find the same premix anywhere in the country, at reasonable prices, and it keeps well. At competitive pricing, I would even tend to think of premix as just another bulk ingredient, like “1 cup flour” vs “1 cup premix FOO”. You can see this with baking powder, chili powder, and curry powder, which you could definitely mix yourself but few bother to do.
ip26··on Flunking my Anthropic interview again
It's never personal

You never screened candidates who couldn’t act their way out of a wet paper bag?

ip26··on Google has eliminated 35% of managers overseeing small teams in past year
And wielding all three at once is the most effective.
ip26··on Branch prediction: Why CPUs can't wait?
It is expensive, but it’s even more expensive not to.
ip26··on Apple's new Processor Trace instrument is incredible
The present opportunity, in my view, is to feed this tracing into the development of superior compilers. This is starting to happen with automated profiling by the compiler, but you can imagine the profiling expanding to an enormous degree, with the compiler tracing the program it is building in great detail.
ip26··on AI is different
For the moment, perhaps it could be jobs that LLMs can’t be trained on. New jobs, niche jobs, secret or undocumented jobs…

It’s a common point now that LLMs don’t seem to be able to apply knowledge about one thing to how a different, unfamiliar thing works. Maybe that will wind up being our edge, for a time.

ip26··on My AI-driven identity crisis
Exactly this sort of thing makes it feel like there could be a "cliff" or "eternal winter of human publishing", where books like his are never again written by humans.

Often when tech comes out that does something better than people, it makes sense for people to stop doing it. But in the case of "books explaining things", AI only learned how to explain things by examining the existing corpus - and there won't be any more human-generated content to continue to learn and evolve from, so the explanatory skills of AI could wind up frozen in 2025.

An alternative would of course be that humans team up with AI to write better books of this sort, and are able to develop new and better ways of explaining things at a more rapid pace as a result.

A relatively recent example that sticks in my mind is how data visualization has improved. Documents from the second half of the 1900's are shockingly bad at data presentation, and the shock is due to how the standard of practice has improved in the last few decades. AI probably wouldn't have figured this out on its own, but is now able to train on many, many examples of good visualization.

ip26··on Yosemite embodies the long war over US national park privatization
It shouldn't be built up like an amusement park, but some level of services seems reasonable. Water, bathrooms, gas for large/remote parks.

If you're advocating even all of that should be removed as well, then we're really talking about converting national parks to wilderness areas.

ip26··on The Ski Rental Problem
The second-hand market is a bit of a mess unless you're comparing among the most popular skis. The depreciation is also brutally steep.

You can avoid the depreciation issue with old skis, but that comes with other problems.

ip26··on Cops say criminals use a Google Pixel with GrapheneOS – I say that's freedom
… three felonies…
ip26··on Cops say criminals use a Google Pixel with GrapheneOS – I say that's freedom
Profiling isn’t taken as evidence of criminal intentions. It can be as simple as, you do random searches, but you increase the frequency of random searches for people who match multiple behavioral criteria empirically associated with criminals.
ip26··on Cops say criminals use a Google Pixel with GrapheneOS – I say that's freedom
“The average person commits felonies a day” is probably a better way to communicate this idea.
ip26··on Nobody knows how to build with AI yet
I do agree… perhaps the thing to do is write fragments of the program, like the start and end, asking it to complete the middle. If you have precisely described how the output will be printed, for example, then you have essentially formally specified how the data should be organized…
ip26··on AI capex is so big that it's affecting economic statistics
including sunscreen as a treatment for lifestyle diseases

HN, where "going outside" is considered a lifestyle.

ip26··on Marathon fusion claims to invent alchemy, making 5000 kgs gold per gigawatt
Given he left Princeton early and is now supposedly synthesizing gold merely two years later, the narrative seems in his favor.
ip26··on Reflections on OpenAI
At some level of raw hours, your health and personal relationships outside work both begin to wither, because there are only 24 hours in a day. That doesn’t always cause burnout, but it provides high contrast - what you are sacrificing.
ip26··on Bill Atkinson has died
If there was a link, I would be thinking about all the superfund sites in Silicon Valley, pondering the manufacture of the Apple II, demographic clustering, or whether there was an unusually strong smoking culture at young Apple Computer, rather than some unique mental stress of the job.
ip26··on Did "Big Oil" Sell Us on a Recycling Scam?
Lead battery recycling is pretty incredible, with allegedly 80% of the lead in a new battery being recycled lead.
ip26··on How Ukraine’s killer drones are beating Russian jamming
I don’t really want to argue this side, but is it that different from a smart bomb or guided missile? A human is in the loop; the human issued the coordinates of the target to the delivery vehicle.

That kind of operation seems extremely different from a stationary turret or patrol robot with standing orders to shoot upon arbitrary targets at any time it decides to.

ip26··on What If We Had Bigger Brains? Imagining Minds Beyond Ours
“printf” is an n-bit integer already. All strings are also numbers.
ip26··on What If We Had Bigger Brains? Imagining Minds Beyond Ours
The language doesn’t just map to English, it allows high level concepts to be expressed tersely. I would bet it’s much easier for an LLM to generate python doing complex things than to generate assembly doing the same. One very simple reason is the context window.

In other words, I figure these models can benefit from layers of abstraction just like we do.

ip26··on Find Your People
One single speech might not do it. However, a short anecdote. Several years ago, while sitting through a training on first aid and cpr on infants, I was bored out of my mind and complaining internally about how this was a waste of time, yet another rehash of common sense. When suddenly it hit me - no, this was not common sense. I had been taught these things repeatedly ever since I was twelve. I didn’t retain all of it after the first time, the second, or the third. But eventually, it was just common sense to me. I couldn’t always tell you where I learned it, or describe the textbook medical guidance. Why is it done precisely this way, who knows, but obviously this exact grip is how you hold the infant while delivering back blows.
ip26··on Find Your People
A wide perspective is good, but that is orthogonal to what they experience as normal. You can select for a good, healthy normal AND provide a wide experience.

Your child doesn’t have to attend a school where educational attainment isn’t valued, to understand that perspective exists.

Their “normal” will strongly influence their choices. For example, if you wanted your child to attend college, I would argue the single best way to ensure they do is to enroll them in a high school where 90%+ of the student body later goes on to college.

ip26··on Find Your People
As categorical guidance, that seems like a problematic thing to teach. First, parents don’t know everything, but they know their kids and have witnessed their journey, so they have a unique perspective to offer. Second, parents will begin to (justifiably!!) grow suspicious of your institution and develop resentment, which is a serious structural problem.

Hopefully what you mean is something like you teach them to think critically about their parents advice as one input among many, understanding where the advice comes from and its inherent strengths and flaws.

← PreviousPage 5 of 34Next →