thelittleone

12 ms·

31.
by thelittleone 2y ago
As a youngster in the 70s (before lead sas a known carcinogen) I had a piece of lead in my lego bag (no idea how it came to be there). I thought that soft metal was neat and bit it more than once. Fortunately I still have all my hair though
32.
by thelittleone 2y ago
Very cool. Though might want to increase contrast on diagrams, for example here https://pico.sh/tuns
33.
by thelittleone 2y ago
Have you had a chance to compare results from MinerU vs LLM such a Gemini 2.0 or anthropic's native PDF tool?
34.
by thelittleone 2y ago
Interesting, although would be great to see some comparative results, e.g., with and without the html alt tag approach.
35.
by thelittleone 2y ago
How about building a tool which indexes ocr chunks / tokens and a confidence grading. Setting a tolerance level and defining actions where the token or chunk (s) fall below that level. Actions could include could include automated veri
36.
by thelittleone 2y ago
Text from diagrams can be useful in LLMs. For example an LLM can understand a flow charts decision making shapes etc, but without text it could misinterpret information. I process a bunch of PDFs including procedures. Diagrams are concerted
37.
by thelittleone 2y ago
Could that be true and at the same time a 'vulnerability' exists that megacorp is party to?
38.
by thelittleone 2y ago
Anthropic has a beta endpoint for PDFs which has produced impressive results for me with long and complex PDFs (tables, charts etc).
39.
by thelittleone 2y ago
1. Work for megacorp 2. Megacorp CEOs gloat about forthcoming mass firings of engineers 3. Pay taxes as always 4. Taxes used to fund megacorp (stargate) 5. Megacorp fires me. The bitter irony.
40.
by thelittleone 2y ago
I sure empathize.... our AI is fussy and rigid... pedantic even.
41.
by thelittleone 2y ago
Agree with you. Time is a huge factor. Another is determination.
42.
by thelittleone 2y ago
No need for sorry, but I appreciate it. I had no intention of implying medicine is simple. Yet the point remains. Doctors cannot read all cases nor memorize all those they do. Unless they have skilled researchers standing by, relevant cases
43.
by thelittleone 2y ago
I've witnessed 4 medical emergencies, with 3 being life-threatening. Each case at different hospitals, and each incorrectly diagnosed. The most dramatic was in 2009. My wife was admitted to emergency at 22 weeks pregnant with severe cr
44.
by thelittleone 2y ago
Please clarify. Do you suggest that because a community is small it doesn't matter?
45.
by thelittleone 2y ago
How to achive total pervasive surveillance? One step at a time where each step is not quite too much to cause rioting and revolution. Outrage has a very short attention span.
46.
by thelittleone 2y ago
Much appreciated. For others interested in BM25 for the use case above, I found this thread informative. https://news.ycombinator.com/item?id=41034297
47.
by thelittleone 2y ago
Not the parent, but we process PDFs with text, tables, diagrams. Works well if the schema is properly defined.
48.
by thelittleone 2y ago
Fixed size chunks is holding back a bunch of RAG projects on my backlog. Will be extremely pleased if this semantic chunking solves the issue. Currently we're getting around an 78-82% success on fixed size chunked RAG which is far too
49.
by thelittleone 2y ago
Those patents wont do much to protect them from competitors abroad.
50.
by thelittleone 2y ago
I find it similarly fascinating. Take for example neuralink. If you consider that interface 10 years, or further 1000 years out in the future, it's likely we will have a direct, thought-based human computer interface. Which is interest
51.
by thelittleone 2y ago
This assumes that prompts do not evolve to the point where grandma can mutter some words to AI that produces an app that solves a problem. Prompts are an art form and a friction point to great results. Was only some months before reasoning
52.
by thelittleone 2y ago
Not only did the megacorp CEO's drop the ball on AI... we've got them gloating over widespread firing of engineers due to AI and then quotes like "I'm good for my $80B" like its his own personal money bag. And now t
53.
by thelittleone 2y ago
I'm not disagreeing, but perhaps during the execution of that project, something far more valuable than next token predictors is discovered. The cost of not discovering that may be far greater, particularly if one's adversaries di
54.
by thelittleone 2y ago
Sadly status quo of the vast majority of enterprise software companies, at least after growth starts to flatten out.
55.
by thelittleone 2y ago
Perhaps he should also talk about the timeline for AI replacing the executive team.
56.
by thelittleone 2y ago
Couldn't we use an AI model trained on historical text data (up to today) to predict likely events for tomorrow? Taking this further, a sufficiently advanced AI system could potentially analyze human-generated text up to any given poin
57.
by thelittleone 2y ago
Interesting. I'm fairly new to the field, but would this configuration help reduce the cost of logging security events from multiple zones/regions/providers to a colocated cluster?
58.
by thelittleone 2y ago
The mid-nineties was a treat for phreaking and hacking.
59.
by thelittleone 2y ago
Peesonally im a Typing Mind user but it got too slow and buggy with long cbaglts. Ended up with boltai which is a natice mac app and found it very good after months of heavy use. I think it could also improve navigation coloring or iconogra
60.
by thelittleone 2y ago
An open source WAF at that.

More ›