raphman

11 ms·

121.
GPT-5 Is a Terrible Storyteller – and That's an AI Safety Problem(christoph-heilig.de)
6 points by raphman 1y ago|0 comments
122.
by raphman 1y ago
Better link: https://www.heise.de/en/news/Paypal-German-banks-apparently-...
123.
by raphman 1y ago
Huh? This user does not look like a bot account at all to me. The account was created in 2016, and the user regularly comments on technical posts - sometimes encouraging developers and providing useful linls. On the other hand, they seem to
124.
by raphman 1y ago
Am I getting that right that your source for this claim is some random person on Twitter who claims to have read an unnamed engineering paper ten years ago? I am no expert on this topic but the first reasonably sciency and recent paper I fo
125.
by raphman 1y ago
Yeah, it's not a wiki at all However, in my experience, many people just know the term 'wiki' as a short form for Wikipedia - and for many, Wikipedia is the only encyclopedia they know. So, I guess that the Deepwiki authors
126.
by raphman 1y ago
The ruling (in German) is available online.¹ The supreme court actually has rejected most of Axel Springer's² arguments and limited the revision to a single question (AFAICT, IANAL): Are the news websites in question some kind of compu
127.
by raphman 1y ago
Actually, the German copyright law explicitly allows people with limited vision or with limited reading ability (not defined precisely in the law) to use assistive technology in order to transfer a work into a more accessible version¹. So,
128.
by raphman 1y ago
Thanks a lot for the explanation and link. I had read the original papers a long time ago, and was not aware of the more recent discussions. That said, I just read a few of the critical papers, and it seems that even Gignac and others do no
129.
by raphman 1y ago
Huh - what do you mean? I just checked again, and this is IMHO exactly what Kruger and Dunning reported: From the abstract: "Paradoxically, improving the skills of participants, and thus increasing their metacognitive competence, helpe
130.
by raphman 1y ago
I have been teaching CS at German universities for close to two decades now. > AFAIK this was because under the old system, there had been too many students registering for exams despite being woefully unprepared, and then predictably fa
131.
by raphman 1y ago
Gergely Orosz' newsletter contained some background on Oxide in 2024: https://newsletter.pragmaticengineer.com/p/oxide
132.
by raphman 1y ago
Nice user experience (though I'd hate waiting so long for the first response). One of my major fears about always-available AI agents is that they will (and already do) infiltrate and disrupt communication in families. In a dystopian
133.
by raphman 1y ago
While I am personally opposed to software patents, I'd argue that the JPEG XS patent holders [1] are not 'patent trolls' in any meaningful sense of the word. While I have no personal experience on that topic, I'd assume
134.
by raphman 1y ago
Very cool - That's nearly exactly what I need for a research project. FWIW, there's also the non-free JPEG-XS standard [1] which also claims very low latency [2] and might be a safer choice for commercial projects, given that ther
135.
by raphman 1y ago
"A Flock of Pigeons is Turing-Complete"
136.
by raphman 1y ago
"Broken" is a spectrum. Adding rapid-fire AI exacerbates the existing problems and makes it harder to fix them.
137.
by raphman 1y ago
> before wrapping in some cling film and pressing it into a patty Why do you wrap it? Couldn't you also form the burger patty without the cling film?
138.
by raphman 1y ago
source for "they do"?
139.
by raphman 1y ago
Not really. Landing gear switches are above the throttle between the screens¹, fuel cutoff switches are below it². ¹) https://youtu.be/RbmFmWqqq0c?t=19 ²) https://youtu.be/33hG9-BCJVQ?t=5 (I'm not an e
140.
by raphman 1y ago
Here's also a video showing operation of the switches: https://www.youtube.com/watch?v=33hG9-BCJVQ
141.
by raphman 1y ago
Dirk Pässler (carbon-drawdown.de) did some back-of-the-napkin calculations for their experiments (enhanced weathering of basalt dust on cropland): > In the end we calculated the actual emissions to be 44,6 tons CO₂, due to the fact that
142.
by raphman 1y ago
A recent paper on this topic (same general message as on the linked website): > We found compelling evidence that honey bee introductions indirectly decrease pollination by reducing nectar and pollen availability and competitively exclud
143.
by raphman 1y ago
Yeah, in Germany pretty much all cans and bottles require a deposit (single-use plastic bottles: 0.25 €) and every shop selling cans/bottles with deposit is required to take them and similar bottles back. Most supermarkets have a rever
144.
by raphman 1y ago
It seems you have the option to run the tools yourself (with an OpenAI API key). The cloud version is for convenience. I agree that a privacy/usage policy is necessary.
145.
by raphman 1y ago
Why and how?
146.
by raphman 1y ago
Thanks! Sorry - I didn't realize that the paper in the ACM DL is not open-access.
147.
by raphman 1y ago
FWIW, I wrote a small paper on this general topic a few years ago, collecting earlier work and own ideas. "Files as Directories: Some Thoughts on Accessing Structured Data within Files" https://dl.acm.org/doi/
148.
by raphman 1y ago
a) because it is a general problem, observed time and again b) I don't see how calling them out by name in this forum might be helpful c) I don't know the whole story - maybe some of these people actually raised concerns or wanted
149.
by raphman 1y ago
Apparently, none of the 21 people mentioned in the acknowledgments questioned the source of the dataset. One of them also wrote a quite popular Twitter thread about the research. When notified of the recent events, he curtly replied that &q
150.
by raphman 1y ago
Thanks. On Twitter, Ethan Mollick seems to imply that Robert Palgrave might be the scientist that triggered the investigation.

More ›