ramraj07

Why blame just the journals when every other system also disintivizes the same.

Do you want issues of Nature and cell to be replication studies? As a reader even from within the field, im not interested in browsing through negative studies. It'll be great if I can look them up when needed but im not looking forward to email ToC alerts filled with them.

Also who's funding you for replication work? Do you know the pressure you have in tenure track to have a consistent thesis on what you work on?

Literally every single know that designs academia is tuned to not incentivize what you complain about. Its not just journals being picky.

Also the people committing fraud aren't ones who will say "gosh I will replicate things now!" Replicating work is far more difficult than a lot of original work.

Not everyone knows who geohot is and even if they do they may not see the url handle. They may (like me) think why is a glorified shower thought tweet on top of HN.

They may not know that this dude was an anti-masker (with nuance) for example. This could really make them decide not to even spend too much time thinking about the passage which in theory is profound for 10 seconds but no further.

As much as ad hominem attacks are not great approaches, the one scenario I feel it's justified is in cases like this.

Hey at least its not ATLAS!

Anyone who's used both Claude and ChatGPT will instantly agree what is better by a large margin. Theres maybe a brand recognition long tail but its more likely theyre the rare occasional users who use the free tier. Thus ChatGPT is becoming the shitty free AI app while Claude is what you use to get real work done. Time (in months) will tell yow this will go.

Avoiding Doing something that could cause job loss has never been and will never be a productive ideal in any non conservative non regressive society. What should we do? Not innovate on AI and let other countries make the models that will kill the jobs two months later instead?

Its definitely against terms. The claude code oauth token is only supposed to be used with claude code. I hope no one gets their claude account banned trying to use this.

To those who use AI to get real work done in real products we build, we very much appreciate the value of each token given how much operational overhead it offsets. A bubble pop, if one does indeed happen, would at best be as disruptive as the dot-com bust.

Star Trek only got a good society after discovering FTL and existence of all manner of alien societies. And even after that Star Treks story motivations on why we turned good sound quite implausible given what we know about human nature and history. No effing way it will ever happen even if we discover aliens. Its just a wishful fever dream.

Great post indeed but let me ask you, put yourself in the LLM shoes. Now instead of reading through coherent lines of code that is exclusively about solving problems, you now have random characters before every line that mean something (because the presence of the edit tool implies it) but not about your actual problem. Do you reckon the LLM will be distracted a little bit? The benchmark deliberately sidestep the actual intelligence of the model on the task at hand, so while the author feels successful at their subtask its very possible they've failed at the war. This seems to be the beauty of AI engineering. The smarter you think you are about something the bigger the fall.

Why do we need a hash for every line. Why cant we mark every fifth line (or get smarter and calculate entropy of lines and jump longer for empty boilerplate)? I feel adding a random 3 char header to every line while making the edit tool smarter will make the overall understandability of the content dumber.

The experience in the claude app is fine.

And blue origin has no delays because they don't even launch anything.

Agreed, every line you ship,whether you wrote it or not, you are responsible. In that regard, while I write a lot of code completely with AI, I still endeavor to keep the lines as minimal as possible. This means you never write both the main code and the tests using AI. Id rather have no tests than AI tests (we have QA team writing that up). This kinda works.

If your MEDIAN is biased, and there are 10+ inputs, the data is fundamentally garbage and biased.

They literally answer that in the post. They started with a single instance and realized sharding the existing tables will be too much work (they'll slowly migrate to new tables instead).

It has one piece of useful info: their main data store even for 800M users is a single instance of postgres (for writes) without sharding.

Dont even need that podcast. Anyone who's done real research in any field, if they try to sit down and read the paper (if you can call it that), can immediately see that its just self aggrandizing, unscientific, biased garbage written by people who think theyre way smarter than they are. Not unlike wolframs new grand unified theory but even worse maybe.

Unfortunately, its also being used by a lot of people who also think theyre smarter than they are to confirm their pre-existing biases with bad research.

Im not saying ChatGPT doesnt make people stupid. It very well might (my hypothesis is that it just accelerates cognition change; decline for many, incline for some). But this garbage is not how you prove it.

I have foregone our take home for exceptional candidates, but let me ask you, do you also demand compensation for in person or zoom call 1-1 interviews? Surely thats the same time of your life.

I recently pasted an error I found into claude code and asked who broke this. It found the commit and also found that someone else had fixed it in their branch.

You should use claude code.

Yes, but the immediate equivalent scenario to me is how people treated other people as slaves merely using them like machines. Sure you got use out of them, but was that the best use?

Its not clear to me that LLMs hallucinating is because of they are extrapolating beyond their training data. Is that proven? Or are you extrapolating?

Even acknowledging it is interpolation, models can extrapolate slightly without making things up, within the range where the model still applies. Whos to say what this range is for an LLM operating in thousand dimensional space? As far as I can tell the main limiters to LLM creativity are guardrails we put in place for safety and usefulness.

And what exactly is your proof that human ingenuity is not just pattern matching. Im sure a hypothesis can be put that fire was discovered by just adding up all known facts the people of those times knew and stumbling on something that put it all together. Sounds like knowledge remix + slight extrapolating to me.

The fundamental idea that modern LLMs can only ever remix, even if its technically true (doubt), in my opinion only says to me that all knowledge is only ever a remix, perhaps even mathematically so. Anyone who still keeps implying these are statistical parrots or whatever is just going to regret these decisions in the future.

E2ee meaning what? Im assuming for compliance reasons any company needs to be able to access all of their internal chats?

Not coding agents but we do a lot of work trying to find the best tools, and the result is always that the simplest possible general tool that can get the job done always beats a suite of complicated tools and rules on how to use them.

I literally have this monitor already and these pixels are humongous. Even at 3 feet away. Also the viewing angle degradation is too much, so much so that it irritates to look at the edge of the screen from the center. A very poor monitor indeed.

The bottom 50% has indeed gotten better (kinda sure but not everywhere) but technological progress is not guaranteeing they'll save everyone _anymore_. Fewer nations care about climate change and taking care of the poorer parts of the world (hope you dont expect me to show why too).

The billion is my number. Almost all scientists and officials underestimate deliberately and naively death tolls. I have a metric that whatever number the new york times publishes for an earthquake on day one, you should multiply it by 100 for the eventual total count. No one estimated a million plus deaths from covid in the US. The moment I read the first comprehensive report from China I assumed 1% of the population will die. Unfortunately in places like India that actually was true. So yeah I do think at the minimum a billion people will die due to climate change in the next 50 years, and billions more will suffer in unthinkable ways. Thats _my_ estimate.

Part of humanity will be sure. But what about the rest?

Climate change will not kill humanity off, but its likely to cause suffering we haven’t seen since WWII (or worse).

Unfortunately I've seen a glimpse of how the bottom 50% of people (not in a developed country but globally) get by today. If one doesnt care for their suffering and their lives, its easy to confirm that the society on average (not on median) will be more prosperous. But that will more likely manifest through a few hundred trillionaires living in space and a few ten thousand billionaires who serve them with their services.

The bottom billions will likely just starve, move around desperately due to war, famine, fire and flood, turned away at closed borders, and who knows what new type of cruelty that will bubble up in the future. What do we tell them?

You want a spray that converts dog poo into dog diarrhea, and you think thats better?

The purpose of these reviewers is to flag the bug to you. You still need to read the code around and see if its valid and serious and worth a fix. Why does it matter if it then says the opposite after the fix? Did it even happen often or is this just an anecdote of a one time thing?