Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
luma
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
13 ms
·
181.
▲
by
luma
2y ago
I’d contest the statement that humans don’t need to be trained to use a calculator. It certainly isn’t instinctive behavior.
182.
▲
by
luma
2y ago
They can do exactly that, it's called Tool Use and nearly all modern models can handle it. For example, I have a consumer GPU that can run a R1 Qwen distill, which, when prompted for a large multiplication, will elect to write a pytho
183.
▲
by
luma
2y ago
> why would anyone pay for this? Presumably, they won't as this is still a tech demo. One can take this simple demonstration and think about some future use cases that aren't too different. How far away is something that'
184.
▲
by
luma
2y ago
This effect has been recently studied: https://www.emergent-values.ai/ They don’t directly say it quite like this, instead letting the data tell a clear story: across vendors and models and architecture and training sets, t
185.
▲
by
luma
2y ago
I've often heard this advice, and while I acknowledge the financial sense it makes, it also puts the new user into a hard place. Here's the problem: the advantages of "good" tools disproportionately impact the new user.
186.
▲
by
luma
2y ago
This is a stunning paper and I’m surprised it didn’t see more traction here. I don’t expect that everyone is going to like the implied outcomes: as these machines get smarter, they’re all becoming more politically liberal and harder to conv
187.
▲
by
luma
2y ago
Waymo is level 4 while FSD is level 3. Waymo won't drive in the snow, because they know they cannot fully automate it safely in all situations. FSD will yolo any situation and just suddenly hand you the wheel whenever it likes, resul
188.
▲
by
luma
2y ago
Ground truthing does in fact mean something different in ML but I’m not clear on how that is troubling.
189.
▲
by
luma
2y ago
Weirdly, they seem more concerned for Tesla the company than they are for themselves.
190.
▲
by
luma
2y ago
Specifically, SAE level 3 should be explicitly prohibited. Humans have proven, over and over, that they can’t handle that level of alertness while also not driving.
191.
▲
by
luma
2y ago
They also created their own scroll, cooked it, and scanned it to create a ground truth dataset.
192.
▲
by
luma
2y ago
None of the orgs you listed provide direct "lawyers for hire" style legal services in the United States. They'll have partner orgs which are lead by lawyers, they'll have international arms with international lawyers, t
193.
▲
by
luma
2y ago
Lawyers have organized their licensing so well that they are the ONLY profession in America not being overrun by PE consolidation because they had to foresight to not allow it. The capital class was eventually going to run up against them
194.
▲
by
luma
2y ago
I think it's largely because SV tech bros have almost zero intersection with enterprise compute as it's used in the rest of the world. PG sees nothing but Apple laptops around him, because everyone holding them has never worked f
195.
▲
by
luma
2y ago
Interesting test and a topic I'm very interested in, but the whole product selection here appears to be very EU-centric. In the states I've been using Weld-On 4 with some success for PETG-to-PETG bonds in high-stress applications:
196.
▲
by
luma
2y ago
That's only a problem if you presume that Google cannot figure out a way to monetize being the second brain that you offload a lot of cognitive tasks to. Hey google I'm hungry, ok how does Pizza sound? Great, make it so. OK, send
197.
▲
by
luma
2y ago
I really wish I could get a dedicated handheld graphing calculator that hasn't had most of the functionality removed due to standardized testing. We live in an age of miracles but having a decent solver on a dedicated device is appare
198.
▲
by
luma
2y ago
Depends on where you live. In most jurisdictions in the US it's generally OK for a homeowner to perform "like for like" replacement work on their own residence without a license or inspection. This means you can replace an
199.
▲
by
luma
2y ago
Not the fella you asked but let me offer some wisdom: smart switches are a lot easier to live with than smart lights. If you also want color control, HA can do a decent job of making smart switches work well with smart lights. The core pro
200.
▲
by
luma
2y ago
Same as every other academic: publish or perish. Collect more citations and note them in your next grant proposal. The modern academy is fucked, the whole system is being gamed and very little is being done about the mess created. The aut
201.
▲
by
luma
2y ago
So how is it then that none of the other models behave in this way? Why is it just Deepseek?
202.
▲
by
luma
2y ago
Use OAI to create synthetic data for your training, which is clearly what they are doing. This is why their models claim to be ChatGPT when asked.
203.
▲
by
luma
2y ago
This is clearly what is happening. Deepseek can train on o1 generated synthetic data and generate a very capable and small model. This requires that somebody build an o1 and make it available via API first.
204.
▲
by
luma
2y ago
Working as intended! The arrs make everything a lot easier.
205.
▲
by
luma
2y ago
I could also have it show me one set of data on my secured machine, but a completely different filesystem + data if stolen and run on some other system, or booted under duress, etc. This seems like a neat feature for some weird use cases.
206.
▲
by
luma
2y ago
The mk4 and mk3 are vastly different machines. If you want to compare the P1S, do it against a contemporary machine. Of course a machine released several years after the mk3 is faster.
207.
▲
by
luma
2y ago
Alternately, more Americans are aware of the threat than you give us credit for.
208.
▲
by
luma
2y ago
Much like Google was eating the lunch of everything in China and the CCP, in response, made it essentially impossible for them to operate. This is not new behavior between the two countries, the only thing new is the direction. US is final
209.
▲
by
luma
2y ago
Spraying your attack into the public with hopes of hitting your target is the polar opposite of responsible. The only "good" part of this is that you were caught in the act before anyone else got hit in the crossfire. In response
210.
▲
by
luma
2y ago
The guy's name is "Hock Tan", it goes all the way to the top. Greedy billionaire trying to squeeze the entire on-prem datacenter industry. Every single one of my VMware customers is either in the process of migrating off or
More ›