kolinko

47 ms·

661.
by kolinko 3y ago
According to the study roughly the same amount of people were optimist as pessimist, if I understand correctly
662.
by kolinko 3y ago
I spotted that as well. But the rest of the study doesn’t mix them. They define realists as people who predicted their future finances correctly, optimists as people who missed (so eg said no change where the reality was their finances got
663.
by kolinko 3y ago
A shoutout to SuperMemo here - one of the oldest SRS out there. Their first version was on Amiga, if I remember correctly, and the newest ones support iOS, use LLMs and all the other tech :)
664.
by kolinko 3y ago
or roots or exchanging it with ground somehow
665.
by kolinko 3y ago
Well, they essentially had the same idea as OpenAI with GPT, just they failed to really build it because transformers weren't invented yet. But their business cases were similar to what we see now with LLMs.
666.
by kolinko 3y ago
As far as I understand, the issue with Cobol specifically is that the code is riddled with gotos and global variables - and untangling all that mess is the real issue, not converting into Java itself. Using traditional algorithms you end up
667.
by kolinko 3y ago
Yes! Also, most people don’t know that the word „Analog”, as in „analog circuits” comes from „analogy”.
668.
by kolinko 3y ago
Programming is not so much about languages as about algorithms, and to a lesser degree - patterns. These don't change much language to language. Also - witnessing the raise of Solidity on Ethereum, and how difficult it was for other la
669.
by kolinko 3y ago
Same in Poland
670.
by kolinko 3y ago
Depends on a country. In Poland, if the light is plain green for cars, it can be also green for pedestrians if you turn left or right, so you need to watch out for pedestrians and other traffic as a driver. If the light is directional left&
671.
by kolinko 3y ago
Are you talking about claude or Gpt4 as well? Anybspecific examples where ChatGPT4 fails for long contexts?
672.
by kolinko 3y ago
8k is very little if you want to add almost any additional data in context, or have a more complicated prompt. Otherwise your knowledge retrieval needs to be almost spot on for llm to provide a proper reply. Ditto with any multi shot prompt
673.
by kolinko 3y ago
True, on the other hand author seems just negative about everything - he expects just single digit satellite launches a year needed, or expects that a permament space station won’t ever happen
674.
by kolinko 3y ago
I’m working on this right now, will be open sourcing soon probably :)
675.
by kolinko 3y ago
Actually, the story doesn't say that it's machines if I can see correctly. It might just as well be other evolved organisms, just not protein/"meat" based.
676.
by kolinko 3y ago
Exhalation is also done as a best audiobook I've listened to - with author's comments (read by himself) interjected between the stories.
677.
by kolinko 3y ago
It writes a function and then you provide it to an interpreter which does the calculation output on which gpt proceeds to do the rest. That’s how langchain works, chatgpt plugins and gpt function calling. It has proven to be pretty robust -
678.
by kolinko 3y ago
LLMs by themselves don’t learn from past past mistakes, but you could cycle inference steps and fine tuning/retraining steps. Also, you can store failed attempts and lessons learned in context.
679.
by kolinko 3y ago
The risk/scenario of singularity is that there will be just one winner and they will be able to prevent everyone else from building their own agi
680.
by kolinko 3y ago
700 out of 770 employees already signed an open letter saying they will consider changing jobs.
681.
by kolinko 3y ago
Yes, it's serving differently depending on IP (or browser preference?). In Poland I see "ousting" as well, but Archive.ph and Archive.org both have "ouster": https://archive.ph/c1kr0
682.
by kolinko 3y ago
I've been using Safari on MacOS for the last 3-4 years (switching from Chrome), and I can't remember a single site that wouldn't work properly. With one notable exception of Google Meet, which consistently tended to fail on a
683.
by kolinko 3y ago
The source for that ( https://archive.ph/OONbb - WSJ), as far as I can understand, made no claim that MS owns IP to GPT, only that they have access to it's weights and code.
684.
by kolinko 3y ago
I think they were exprimenting with a more advanced 3.5 version for non-paying users - it was showing to some people as an option between gpt-3.5 and gpt4
685.
by kolinko 3y ago
theverge and theinformation have tip lines/e-mails, and you can message them, or write to one of the reporters writing about that directly. I would trust them to keep the confidentiality
686.
by kolinko 3y ago
That’s a terrible title
687.
by kolinko 3y ago
You haven't been around when Web2.0 and the whole modern internet arrived, were you? You know, all the sites that you consider stable and robust now (Google, YT and everything else) shipping with a Beta sign plastered onto them.
688.
by kolinko 3y ago
It brings a tremendous uncertainity in the direction the OpenAI is heading. They could block out APIs now, or do other kinds of strategic shifts that would affect the users. So far, from my viewpoint, OpenAI was perfectly as a product - bot
689.
by kolinko 3y ago
From Tom Perkins’s biography - after serving on boards both big private companies and non-profjts, he said that non profits were much worse. His theory was that with no money on the stake it’s all about egos, and they cause weird situations
690.
by kolinko 3y ago
That is what's happening. When you buy a plane ticket, in many airlines, you have an option to pay extra to fly on SAF.

More ›