HackerTrans
TopNewTrendsCommentsPastAskShowJobs

vlan121

73 karmajoined 5 lat temu

Submissions

Ads in ChatGPT

ads.openai.com
4 points·by vlan121·8 dni temu·1 comments

Claudes newest update is putting a worse model to default

github.com
1 points·by vlan121·10 dni temu·3 comments

[untitled]

2 points·by vlan121·8 miesięcy temu·0 comments

Losing Trust in the Internet

13 points·by vlan121·9 miesięcy temu·8 comments

Humanties Last War

1 points·by vlan121·9 miesięcy temu·1 comments

comments

vlan121
·5 dni temu·discuss
The ai training data flag
vlan121
·10 dni temu·discuss
Thanks!
vlan121
·10 dni temu·discuss
Changelog: https://github.com/anthropics/claude-code/releases

Performance is worse than Opus 4.8: https://www.anthropic.com/news/claude-sonnet-5
vlan121
·10 dni temu·discuss
This is one of the core principles of How To Win Friends And Influence People (Dale Carnegie). Since Ive read I tried to apply this rule. It works.
vlan121
·21 dni temu·discuss
Why a shell company in Stockholm?
vlan121
·27 dni temu·discuss
Maybe this is the line, we'll hit eventually. Maybe the models become smarter, but the context will sit.
vlan121
·w zeszłym miesiącu·discuss
Reversly viewed ones should ask with what intend the study should be like this. And for obvious reasons it sounds like monetary-nature.
vlan121
·8 miesięcy temu·discuss
I was unable to find the paper. I'm still wondering, if it is a cross-over experiment, as:

> The circles were in different positions at each room in the maze, but the bees still learned over varying amounts of time to fly toward the short flash of light associated with the sweet food.

Do not state, if the light suddenly changed in the rooms. If not, other factors might come into place.
vlan121
·9 miesięcy temu·discuss
> I mostly lucked out. 2-3 accounts surfaced over the many things I saw and they seemed to prove they knew things. For example, one of these accounts posted things 2-4 years before an event would occur that ended up coming true. As if they knew ahead of time what would happen, as if they were part of intelligence or government or who knows what. A different account created software that predicted the future and many of the things they posted ended up coming true too.

We have to be cautious here, it's drifting towards conspiracy theories, my post is not innocent in this regard. But apart from that, I have to agree with you.

> In many areas, I don't fully understand. I use a simple trick that worked well: [phrase-redacted]. Intellectually honest people who aren't trying to trick anyone speak the truth. They also speak it with simple words so there's no possibility of misinterpretation.

So not taking a look at the chain of thoughs but treating each sentence differently, makes sense.

> This might be the most effective trick. Many times I caught people say something which is their opinion, and which is false based on data I have, that is actually an attempt to express a fact they know in a way that will make them seem more important, or give them more attention, or I don't know what.

Seems like philosophical razor

>On the other hand, a social credit system may be a distraction from an actual credit system already in place: a money system.

That is very true.
vlan121
·9 miesięcy temu·discuss
> I follow content from accounts that seem to consistently post facts. How do you really distinguish that? Sourcing is a way to proof your facts. But creating a net of sources is in times of AI not hard. If there are timestamps one factor could be time stamps, and I really found sources that all have been published within a week or a month making it shady.

>I try to understand what is actually being stated instead of what is implied; facts vs opinion That requires that you understand fully to implication, how do you do this in areas that you don't have strong background knowledge?

>I wouldn't be surprised if falsity starts to get penalized. Elon said Grok will start to rank content based on whether its true. That would be a great step.

> Also built a system that prevents posting things that are made up. Maybe posting lies should carry a penalty, where you deposit something to lose if you told a lie. We must be careful here that a social credit system is not created through the back door.
vlan121
·9 miesięcy temu·discuss
SocialFocus offers so much more than just shorts :-)
vlan121
·9 miesięcy temu·discuss
Security 101: * Kerckhoff Principle of Open Design Security of a mechanism should not depend on the secrecy of its design or implementation.
vlan121
·10 miesięcy temu·discuss
I dont know about this. You can look up a video on YouTube but when you search for valid sources I just find magazines I never heard of. Is there a valid source?