fumeux_fume

5 ms·

61.
by fumeux_fume 1y ago
I like that OpenAI is drawing a clear line on what “hallucination” means, giving examples, and showing practical steps for addressing them. The post isn’t groundbreaking, but it helps set the tone for how we talk about hallucinations. What
62.
by fumeux_fume 1y ago
In the article, OpenAI defines hallucinations as "plausible but false statements generated by language models." So clearly it's not all that LLMs know how to do. I don't think Parsons is working from a useful or widely a
63.
by fumeux_fume 1y ago
> Hallucination is all an LLM does. I wish people who take this stance would seriously reconsider their take on how hallucinations are defined and how unhelpful it is to conflate hallucination with generation from a probability distribut
64.
by fumeux_fume 1y ago
The author's inner PM comes out here and makes some wild claims. Calibration is something we can do with traditional, classification models, but not with most off-the-shelf LLMs. Even if you devised a way to determine if the LLM's
65.
by fumeux_fume 1y ago
It'll force a response that begins with an open bracket. So if you might need a response with a tool call that doesn't start with "{", then it might not fit your workflow.
66.
by fumeux_fume 1y ago
We all have allergies. I'm allergic to 1000 line SQL queries which include functions that are only usable for a specific flavor and version of SQL.
67.
by fumeux_fume 1y ago
In the same talk, Mark acknowledges that "for data science workflows, database systems are frustrating and slow." Granted DuckDB is an attempt to fix that, most data scientists don't get to choose what database the data is st
68.
by fumeux_fume 1y ago
Very common struggle, but a great way to prevent that is prefilling the assistant response with "{" or as much JSON output as you're going to know ahead of time like '{"response": ['
69.
by fumeux_fume 1y ago
I will always remember when my Dad bought a vertically flat, 27in Trinitron back around 1998. I miss those buttery-smooth pans. Probably my biggest gripe with any modern television is how awful panning or tracking shots look. Similarly, I e
70.
by fumeux_fume 1y ago
The horrible scale system of self-checkouts brought my anxiety to a fever pitch. Any slight adjustment to the bag or moving anything around would literally set off an alarm for "assistance." Still gives me low-key ptsd even though
71.
by fumeux_fume 1y ago
Parent comment is literally praising an experience they had in higher education, but your only takeaway is that it must be facile ridicule of academia.
72.
by fumeux_fume 1y ago
Sometimes there little to zero negativity or criticism and yet, the top post is "I'm surprised by the negativity..." It's disheartening to see Reddit-level manipulation of the comment section on HN, but I accept that shi
73.
by fumeux_fume 1y ago
So using some complicated angle gets you 4% less std dev than just doing it the easy way that everone already does it. Ok.
74.
by fumeux_fume 1y ago
The information you're seeking appears to be left out of the post. My best guess is that a separate embedding model, specifically tuned for document similarly, is used to generate the vectors and then a clustering algorithm is chosen t
75.
by fumeux_fume 1y ago
Slow or fast ultimately matter in the context for which you need to use it. Perhaps these are only myths and fairly tales for an incredibly small subset of people who value execution speed as the highest priority, but choose to use Python f
76.
by fumeux_fume 1y ago
Arch worked just fine for me
77.
by fumeux_fume 1y ago
Wholesale use of and/or dependence on Claude Code feels like a Faustian bargain from a personal and even a business point of view.
78.
by fumeux_fume 1y ago
Thank you! When I was testing out Copilot I was stuck with whatever default LLM was being used. Didn't realize you could switch it out for a non-MS/OpenAI model.
79.
by fumeux_fume 1y ago
I want to use Gemini 2.5 PRO. I was an early tester of Copilot and it was awful.
80.
by fumeux_fume 1y ago
I don't want to leave my IDE
81.
by fumeux_fume 1y ago
I currently use LLMs as a glorified Stack Overflow. If I want to start integrating an LLM like Gemini 2.5 PRO into my IDE (I use Visual Studio Code), whats the best way to do this? I don't want to use a platform like Cursor or Claude C
82.
by fumeux_fume 1y ago
Of course that’s angle they decide to open the article from. That they feel the need to frame these tools using the most grandiose terms bothers me. How does it make you feel?
83.
by fumeux_fume 1y ago
The FTC under the current administration has zero interest in pushing this forward
84.
by fumeux_fume 1y ago
This is a pretty narrow view. A lot of businesses--whose bread and butter (well maybe just the butter) is keeping people locked into subscriptions they don't want--put a large effort in challenging this rule. They would have fought it
85.
by fumeux_fume 1y ago
I wonder if the decision to forgo the review process was a cynical gamble knowing it would be slow-walked to death or if it was done to score quick points with little concern given to how a legal challenge would play out in the future.
86.
by fumeux_fume 1y ago
The Last Supper was also a fresco painted with oils, IIRC. The painting quickly devolved and needed constant restoration.
87.
by fumeux_fume 1y ago
“Wrap your lips around this.” Often gets quoted in our household when serving a drink. RIP MM.
88.
by fumeux_fume 1y ago
ACSM download, DRM-stripping and uploading to Dropbox can be automated with a python script if you buy from the Kobo store. It requires Calibre, but you just need it for the CLI. Another python library is needed for automating the ACSM auth
89.
by fumeux_fume 1y ago
If your fishing net is constructed just right--so as to pick out the many interesting gems in this article, you will be glad to know Gefter has also published a book that revolves around John Wheeler.
90.
by fumeux_fume 1y ago
Lol, a bold claim. It's a rational assumption that any business publishing "academic work" is selling you the upside while omitting or downplaying the downside.

More ›