mgraczyk

13 ms·

181.
by mgraczyk 1y ago
This isn't true in the US. Young people in US consume much more of those things you listed than people over 40 did at the same age. Young people have more purchasing power than previous generations. EDIT: Data from the fed and payroll
182.
by mgraczyk 1y ago
Does that not match your experience? For example 250 base, 750 equity wouldn't be out of the norm
183.
by mgraczyk 1y ago
I read it backwards, 75-90 is equity. I fixed. But I should make it clear that ordinarily I'd agree with the haters, but AI is actually different and so much wealth will be created that many of these unicorns will end up exiting for pl
184.
by mgraczyk 1y ago
Sure, but some of these companies are valued at multiples lower than public companies doing AI. It's reasonable to discount that but foolish to ignore it
185.
by mgraczyk 1y ago
0-10 years experience (wide range) Let's say top 30 by valuation Depending on the startup, 10-25% Very widely I claim the median is around 1M at these top companies. Not the median offer, but the median current employee
186.
by mgraczyk 1y ago
There is no crash if you like what you're doing and are surrounded by people who work hard and feel like they're on the same team
187.
by mgraczyk 1y ago
A typical engineer at a top AI startup makes over 1M/year, mostly equity. You may think that the equity is worthless, but they don't and that's part of why they are willing to work hard
188.
by mgraczyk 1y ago
A lot of people working these jobs are being paid millions per year, it's really not the right group to feel bad for. Also nearly everyone I know who does this loves it
189.
by mgraczyk 1y ago
Do you work at an AI startup? What would the union do to help you?
190.
by mgraczyk 1y ago
I wasn't being serious when I said it, I was using it as an insult for bad work
191.
by mgraczyk 1y ago
I'd like to brag that I got in trouble for saying this to somebody in 2021, before ChatGPT
192.
by mgraczyk 1y ago
What I mean is that clearly processed foods aren't harmful because they are processed. All the correlations go away when you control for basic things like sugar and vegetable consumption. The whole idea that processed foods are bad for
193.
by mgraczyk 1y ago
Why do you believe it's not healthy? Because of the processes food hysteria stuff?
194.
by mgraczyk 1y ago
Probably one person one minute, unless it needs translation in which case potentially unbounded time
195.
by mgraczyk 1y ago
Then why does the EU commission believe they need one for their pages that explain these rules? What invasive telemetry are they using? Are there any websites that don't require these banners? If you allow users to set font size or lan
196.
by mgraczyk 1y ago
The reality is that most people who smoke/vape indoors will lie about it. I've witnessed this hundreds of times from hundreds of people. In every place I have lived I had neighbors who smoked (illegally) and lied about it to my f
197.
by mgraczyk 1y ago
This is actually not true. I just read the European commission's cookie policy. The main reason they need the banner is because they show you full page popups to ask you to take surveys about unrelated topics like climate action. They
198.
by mgraczyk 1y ago
Even EU government websites have horrible intrusive cookie banners. You can't blame ad companies, there are no ads on most sites
199.
by mgraczyk 1y ago
Imagine a world where there is 10x as much wealth and 10x as many hard problems being solved. Suppose there's even a 5% chance of that happening. It's clearly worth doing
200.
by mgraczyk 1y ago
I currently use LLMs to design code with ambiguous requirements. It works quite well and scales better than hiring a team. Not perfect but getting better all the time. The key is to learn how to use them for your use case and to figure out
201.
by mgraczyk 1y ago
Important for HN users in particular to keep in mind: It is possible (and IMO likely) that the article is mostly true and ALSO that software engineering will be almost completed automated within the next few years. Even the most pessimistic
202.
by mgraczyk 1y ago
I'm not talking about fake placards. You can get real ones online very easily via telehealth, or in person from many chiropractor or other sham providers
203.
by mgraczyk 1y ago
You can easily buy handicap placards online. It's actually pretty similar to the dog thing, mostly a way to get out of paying for parking
204.
by mgraczyk 1y ago
Ah you are correct: https://techcrunch.com/2014/06/26/how-can-chromecast-connect... I couldn't remember if it actually shipped in anything, but looks like it did!
205.
by mgraczyk 1y ago
Google explored doing this with devices that connect to each other, specifically Chromecast and their early Google Home devices. I don't think it ever launched but Google did some interesting experiments to test the ultrasonic transfer
206.
by mgraczyk 1y ago
But that view is wrong, the model outputs multiple tokens. The right alternative view is that it's an immutable function from prefixes to a distribution over all possible sequences of tokens less than (context_len - prefix_len). There
207.
by mgraczyk 1y ago
LLMs, considered as a function of input and output, are not immutable at runtime. They create tokens that change the function when it is called again. That breaks most theoretical arguments
208.
by mgraczyk 1y ago
Well no, that's not what anyone is saying. The claim was that it isn't possible in principle for "DAGs" or "immutable architectures" to be intelligent. That statement is confusing some theoretical results that
209.
by mgraczyk 1y ago
Because that's quite a bit more information processing than any human brain
210.
by mgraczyk 1y ago
That doesn't matter, are you familiar with any theoretical results in which the computation is somehow limited in ways that practically matter when the context length is very long? I am not

More ›