leobg

12 ms·

181.
by leobg 1y ago
Hard not to read your view as a textbook example of projection, given your username.
182.
by leobg 1y ago
Ron Reagan, My Father at 100: > Many victims rapidly developed an especially viscous pneumonia and then, as one physician put it, “died struggling to clear their airways of a blood-tinged froth that sometimes gushed from their nose and m
183.
by leobg 1y ago
“Confabulation” is the better term imho (literally: making things up). But I guess OpenAI et al stuck to “hallucination” because it generalizes across text, audio and image generation.
184.
by leobg 1y ago
Clickable: https://data-star.dev/
185.
by leobg 1y ago
To save some people a click: > According to a New York Times review, Stanley's book—a "slim volume"—"breezes across decades and continents" and says that Donald Trump "resembles other purveyors of authoritar
186.
by leobg 1y ago
Interesting. WhisperMemos user here. I used audio journaling before Whisper came out, and I did have a pipeline that would run a transcription through Google Cloud and save the transcript to Evernote. I didn’t actually review the transcript
187.
by leobg 1y ago
I agree. PDFs/papers are no fun on a phone. I heard “book” and thought of fiction / nonfiction.
188.
by leobg 1y ago
BF Skinner had the same idea. You can read it in his book, Walden Two.
189.
by leobg 1y ago
What about book reading requires a large screen? I find my phone much more comfortable to hold than a book. It fits my hand. You only read one sentence at a time anyway. I rather scroll, keeping the current sentence in the middle of the scr
190.
by leobg 1y ago
Yes, exactly. You want to randomize the parts that are irrelevant. For example, if you're classifying news articles, you may want to shorten them anyway. A human would be able to tell what category an article belongs to without reading
191.
by leobg 1y ago
Or from Taylor’s.
192.
by leobg 1y ago
If I have 1,000 labeled examples for a classification task, I’ll expand that into a training dataset using augmentation, and then finetune a small model like RoBERTa. It’s fast, cheap, accurate — and predictable. Others have had success wit
193.
by leobg 1y ago
Did you ever make the effort to ask why he did those things? There’s a possibility that the person who aimed high for team humanity with SpaceX, Tesla and Neuralink was doing exactly the same thing in the examples you named.
194.
by leobg 1y ago
I hope the owner sues them not only for the painting, but also for interest / rent for those ~90 years.
195.
by leobg 1y ago
I had the same problem. All SSL connections are failing. I now run a copyparty server, without SSL, allowing my iPad to stream audiobooks and videos.
196.
by leobg 1y ago
That was my point. Knowledge based on a headline based on a tweet.
197.
by leobg 1y ago
> Psychiatrist and philosopher Thomas Fuchs points out that while chatbots can make people feel understood or cared for, especially in therapy or companionship settings, that sense is just an illusion that can fuel delusions or replace r
198.
by leobg 1y ago
So if I want to discourage AI extraction on my site, I can just set up my server to respond to Gemini requests with an inflated page, and then the Gemini customer is going to pay for those tokens? I don’t see any cap. No de-duplication. No
199.
by leobg 1y ago
That’s the hallmark of socialist countries. It used to be that where in Eastern Germany people would look to government for a solution, those in the West would start a business, or learn a new skill, or read self help books. Nowadays, the s
200.
by leobg 1y ago
„If I Were The Devil“ (1965) https://youtu.be/jnPE8u5ONls
201.
by leobg 1y ago
> based on the headlines they skimmed about Musk’s politics
202.
by leobg 1y ago
Another way to see it is that Flock merely wants to place itself at the top of the crime food chain.
203.
by leobg 1y ago
Less lobotomized and boxed in by RLHF rules. That’s why a 7b base model will “outprose” an 80b instruct model.
204.
by leobg 1y ago
“I bought this car before Adolf became problematic.”
205.
by leobg 1y ago
This is an incredibly sad read.
206.
by leobg 1y ago
I agree with you that, of course, there are people and units in government that may be highly skilled and efficient. From what I see, here in Germany, and I think in the United States as well, is that those are the exception and not the rul
207.
by leobg 1y ago
My take is that government is like a really lazy college student. Goimg to the library to study would be hard, and you’d need vision and motivation to do it. Instead, you take the money given to you by your parents, buy the best textbook th
208.
by leobg 1y ago
They take an exorbitant fee to clean up the mess government created when they outsourced their tech infrastructure to private sector companies preying on dumb government money. That’s the thing with government: They always believe you can d
209.
by leobg 1y ago
This was a great write-up! Didn't you run into Cloudflare blocks? Many sites are using things like browser fingerprinting. I'd imagine this would be an issue with news sites particularly, as many of them will show the full content
210.
by leobg 1y ago
Kokoro is small and fast because all the text -> phoneme conversion is done by “dumb code” and only the phoneme -> sound part is done using a neural net.

More ›