a_bonobo

33 ms·

451.
by a_bonobo 4y ago
How many 'thought leaders', 'public experts', podcasters are purely experts at being experts, with no experience or knowledge behind it? I can think of a few public voices, especially in AI right now, who have nothing to
452.
by a_bonobo 4y ago
You're right, here's some data to support your point: >In 2020, the reported violent crime rate in S.F. was 21% below the average of 20 most populous cities other than San Francisco for which there was full 2020 data available.
453.
by a_bonobo 4y ago
Interestingly, if you have a gun, you're more likely to die of homicide https://www.scientificamerican.com/article/more-guns-do-not-... Quote: >Every time a gun in the home was used in self-defense or legally j
454.
by a_bonobo 4y ago
Very cool solution! My solution was to write a bit of code that writes a CSV, then I used a langchain-based CSV agent. Since that one calls on pandas it effectively has no token limit, but it also has no overview of the data, only what pand
455.
by a_bonobo 4y ago
There's this cool SF novel by Polish author, The Old Axolotl by Jacek Dukaj. Some humans upload their minds into virtual reality game as mankind dies out. The remaining humans live forever, but the novel makes it clear that they stop &
456.
by a_bonobo 4y ago
'Racial homogeneity' is not a thing in biology Furthermore, up until 1869 Hokkaido was a separate kingdom peopled by the Ainu - Japan has a long history of exchanging people with Korea. It's not some fabled walled-off kingdom
457.
by a_bonobo 4y ago
This is a myth - Japan does not collect data on ethnicity/background [1], so we just don't know. Japan has a (to foreigners) invisible and extensive caste system [2] and a long history of looking down at Japanese of Korean backgro
458.
by a_bonobo 4y ago
> like someone telling an elaborate lie just for the fun of knowing that people are falling for it I believe Pynchon is the grand master of telling Shaggy dog stories. Norm MacDonald was a huge fan of Pynchon, it feels like they had&#x2F
459.
by a_bonobo 4y ago
Yes, I fully agree with you! there's no reasoning or intelligence in modern LLMs, but the OpenAI open letter and recent comms strongly implies there is
460.
by a_bonobo 4y ago
In this vein, ChatGPT is a nice way to start thinking about any topic; ask it about its opinion on anything and it will give you the most run-of-the-mill, middle-of-the-road text that is possible, standing for nothing. If you find yourself
461.
by a_bonobo 4y ago
>Is there any Researcher who maintains that LLM models contain Reasoning and intent? That's the funny thing about the (in)famous OpenAI letter; the first sentence kind of does this: >AI systems with human-competitive intelligence
462.
by a_bonobo 4y ago
that's it for me as well; when I was a child, McDonald's was the cheap option. Nowadays it's more in the middle of the price range of where I get my food, which is crazy considering what you're getting. It's cheaper
463.
by a_bonobo 4y ago
If you're a German computer kid millennial you've likely seen 23, a relatively-fictional movie that focuses on Karl Koch's life and death. https://en.wikipedia.org/wiki/23_(film) I've discovered the
464.
by a_bonobo 4y ago
I'd not call any of those rambling messes publications >By far the greatest danger of Artificial Intelligence is that people conclude too early that they understand it. Of course this problem is not limited to the field of AI. Jacqu
465.
by a_bonobo 4y ago
I don't think there is any. It's all this alarmist-rokos-basilisk-phpBB nonsense, no code produced, no actual research. What annoys me more is that with his institute's closeness to the SV crowd, 'real', 'groun
466.
by a_bonobo 4y ago
over the past 5-10 years my field biology has embraced preprints, with some famous papers remaining as preprints (most famously, I'd say, is the BWA-MEM paper). Most preprints I see are of decent quality, certainly higher than what ge
467.
by a_bonobo 4y ago
can we move to post-publication peer review now? Pre-publication peer review, most of the time, marginally improves a manuscript at the cost ~6 months. Some times it's just a place for reviewers to vent off stress and punch down. The r
468.
by a_bonobo 4y ago
Standard GPT models are not very good at comparing DNA-sequences, unfortunately https://www.biorxiv.org/content/10.1101/2023.03.11.532238v1 See Figure 2, DNA alignment is mostly 0 accuracy
469.
by a_bonobo 4y ago
This would also explain the ease at which ChatGPT gets rid of escapes/bad prompts - they have an additional layer that assesses whether the question could be, for example, racist, and then spits out a 'Sorry as a language-model I
470.
by a_bonobo 4y ago
llama's 7b model internally is, for me, on a totally different level quality-wise. Even when explicitly instructed to not make up stuff and just say 'I don't know', it will still go ahead and ramble and invent things. Wh
471.
by a_bonobo 4y ago
I asked ChatGPT: 'what is a man?', and it showed me a picture of a plucked chicken saying, 'Behold! I've brought you a man!'
472.
by a_bonobo 4y ago
>The thing is, we both were not at the same level, quality-wise. My work was always a tad better, in shape and texture, rendering… I always was very sure I wouldn’t loose my job, because I produce slightly better quality. One thing I had
473.
by a_bonobo 4y ago
teddit.net and libredd.it are alternative interfaces that still work (until they shut down the API)
474.
by a_bonobo 4y ago
The big problem of language models is that we can't assess how truthful their statements are, only how grammatically correct they are. I know people are working on mathematically assessing truthfulness of a statement for the next gener
475.
by a_bonobo 4y ago
I've been playing a lot with the langchain Document Loaders - their examples make the results look more impressive than they are, or perhaps that's because our documents are 'special' (heavy in taxonomy/biology). I&
476.
by a_bonobo 4y ago
>what are the known downsides that people in academia already know about? I work in biology with some interface to wildlife managers/'end users' of our biological insights. I was hoping we could use these models as advance
477.
by a_bonobo 4y ago
Cowardice You fired someone? You're to blame. Someone was impacted? Sounds more like an act of nature. Nobody is to blame!
478.
by a_bonobo 4y ago
Asimov is also quite handwavy in how these laws are implemented within the robots, making them out to be written in stone. I haven't encountered an LLM with a premise that cannot be broken out of.
479.
by a_bonobo 4y ago
Nature News is a weird place. They are editorially independent from main Nature, and as such they often make the 'right noises': articles decrying the ever-lasting hunt for impact factors and glam journals, articles hyping up ope
480.
by a_bonobo 4y ago
You're right! The only sentence about India: >An expansion [ of the family from the Gulf of Carpentaria] at this time is consistent with evidence from early genetic studies for gene flow into Australia from India ~4–5 ka21, but more

More ›