Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
daveguy
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
34 ms
·
601.
▲
by
daveguy
1y ago
> Again, the paranoia is just drivel. Well, in this domain (government surveillance), probably not paranoia. https://en.m.wikipedia.org/wiki/Edward_Snowden
602.
▲
by
daveguy
1y ago
Pretty sure just one human can handle it, PhD or not. The AI LLM variety, I don't think any number of them could given their lack of reliability.
603.
▲
by
daveguy
1y ago
Shouldn't "nearly AGI" levels of intelligence in an AI be smart enough to handle "if statement" logic? Altman is running around saying you have a bunch of PhD level experts in your pocket with GPT5. How many PhDs do
604.
▲
by
daveguy
1y ago
They made the mistake of assuming the AI was intelligent. I'm not sure where they got that notion.
605.
▲
by
daveguy
1y ago
FTA - This is after McDonalds cancelled AI in their drive through last year after people getting orders for bacon on ice cream (which can be delicious, tbf) and hundreds of dollars of unwanted nuggets. So... AI isn't even smart enough
606.
▲
by
daveguy
1y ago
I would expect it to be almost tautologically true. It's the easiest to implement, and the most deployed. It has to be the most used, even if a lot of people don't know they're using it. It was the original success of LLMs --
607.
▲
by
daveguy
1y ago
I would be very interested in reading about huge bubbles in railroad, airline, or any other industry. Do you happen to have references (genuinely asking; the original article didn't include any references)? Edits-- Found one: https:&
608.
▲
by
daveguy
1y ago
Okay, but LLMs will produce a thousand lines of code for Hello World. Plus, there is no telling how many bugs you have in that code.
609.
▲
by
daveguy
1y ago
> Why would I sane person keep putting these uncontrollable spike-protein-producing machines in their bodies?! These vaccines produce spike proteins for one expression cycle proportional to their dose. They are far from "uncontroll
610.
▲
by
daveguy
1y ago
That is completely valid by the constitution. The judiciary is clearly still a check. The judiciary striking this down would mean that Congress has one less check on the executive, which is exactly what Project 2025 wants to install an unac
611.
▲
by
daveguy
1y ago
There are millions, if not hundreds of millions (probably less than a billion), of free adults living around the world without subscriptions. Many in the US too (250k is a low estimate) of thousands to a mi, don't require subscriptions
612.
▲
by
daveguy
1y ago
The parent post was an awfully broad sweeping statement that deserved an accurate reply. Obviously we don't "all need subscriptions to stay alive." I guess folks on hn have forgotten the difference between need and want.
613.
▲
by
daveguy
1y ago
No.
614.
▲
by
daveguy
1y ago
1) a well designed test suite does run fast. 2) they very easily could have determined how readily the bot talks about suicide under guise of "character development" or whatever they call that very well known jailbreak. 3) move fa
615.
▲
by
daveguy
1y ago
Yes, Facebook was one of the original shitty actors. They popularized/coined the phrase. It was their motto.
616.
▲
by
daveguy
1y ago
To be fair. "Move Fast and Break Shit" was a garbage motto to begin with. Maybe better to be at least slow enough to realize the consequences of a broken app. If you're moving too fast and breaking too much shit, you won'
617.
▲
by
daveguy
1y ago
Apparently Silicon Valley VC culture is trying to transition from move fast and break things to move fast and break people.
618.
▲
by
daveguy
1y ago
That's because it's not an effort to get rid of waste, fraud, and abuse. It's a scam to steal data and give Republicans something to say.
619.
▲
by
daveguy
1y ago
I'd rather a bunch of move fast and break things short-sighted dipshits didn't make the risk higher, but here we are.
620.
▲
by
daveguy
1y ago
They had the exact same thing last time too. Control over all arms + support of much of the private sector. We told them to go fuck themselves then and we will again.
621.
▲
by
daveguy
1y ago
Maybe the role reversal breaks most of the RLHF training. The training was definitely not done in the context of role reversal, so it could be out of distribution. If so, this is a glimpse of the intelligence of the LLM core without the RL&
622.
▲
by
daveguy
1y ago
Well, they've said they're close over and over again. Maybe that final bit of tech to make AGI a reality will finally ride into existence on the sub-$30k Tesla.
623.
▲
by
daveguy
1y ago
That's what they said in 2016.
624.
▲
by
daveguy
1y ago
I don't, sorry. These are just images that I pulled off this site: https://www.landroverusa.com/vehicles/defender/interior-gall... (probably should have lead with that instead of just dropping direct links)
625.
▲
by
daveguy
1y ago
Okay, but maybe recruiters aren't up on rust iconography. Seems like an ineffective way to communicate a preference (vs it being a shibboleth with the GP). I mean, unless the recruiter is a rust programmer they're going to have a
626.
▲
by
daveguy
1y ago
[0] https://jlr.scene7.com/is/image/jlr/L663_22MY_031_US?wid=480... [1] https://jlr.scene7.com/is/image/jlr/L663_23MY_014_US?wid=480... [2] https://jlr.scene7.com&#x
627.
▲
by
daveguy
1y ago
That makes fundraising easier, by increasing the appearance of authority and therefore coming off as a "better" model. In the elo ratings I'm sure GPT-5 is doing better because of the clear push back against these LLMs as way
628.
▲
by
daveguy
1y ago
Aren't they discontinuing 4.5 in favor of 4.1? I think they already have with the API.
629.
▲
by
daveguy
1y ago
Apparently you do not understand how they chose the two groups. Group identity was not based on a survey or any attribute of the participating individuals. Low and high socio-emotional groups refer to whether the group was shown the low or
630.
▲
by
daveguy
1y ago
Well, it helps that Russia has captured and helped pump propaganda over well more than 50% of US information channels.
More ›