scottwambler.bsky.social
Father, husband, author, and consultant in better ways of working (WoW) and ways of thinking (WoT) around data warehousing, data quality, and AI development. I help data people get better at delivering real value.
56 posts
200 followers
392 following
Prolific Poster
Conversation Starter
comment in response to
post
Keep going, there will be a McDonalds or Starbucks any time now. Or a Denny’s.
comment in response to
post
I’m around 5%. I think there’s opportunity for malicious use of AI to do some nasty stuff, particularly overcoming security of critical infrastructure. Never underestimate the ingenuity of terrorists with new tools.
comment in response to
post
Doesn’t mean they’re not going to try.
comment in response to
post
I would call the role #data archeologist, a data analyst who investigates the (currently) unknown lineage of data.
comment in response to
post
Today years old. Thanks for that. I’m telling mom!
comment in response to
post
After an evening of great food and drink, perhaps too much drink, I made a bit of extra money belly dancing.
comment in response to
post
Happy birthday Millie! America is giving thanks today in recognition of you being such a good dog. May the force be with you, so say we all.
comment in response to
post
Here you go:
metro.co.uk/2024/11/27/u...
comment in response to
post
And yes, feel your pain on this as I have a similar issue. “But it was such a good idea” and “Maybe I will get around to this” thoughts aren’t helping at all.
comment in response to
post
Have you had offers for the domains over the years? If so, hang on to those ones and sell next time you get an offer. If not, unless you intend to actually take up the project I would let them expire.
comment in response to
post
I believe that we used to have this, but that it was slowly whittled away over time by a minority who did it for their own financial benefit. BlueSky is a reset back to those times. Hopefully we don’t mess it up.
comment in response to
post
Yes. The majority of the blog was an overview of how to increase the quality of the hallucinations (better prompting, better data, RAG, ...).
comment in response to
post
Some thoughts:
1. Yes, errors in the process are also a problem.
2. To be exact, it's LLMs that hallucinate (100% of the time, it's the quality of the hallucination that's the issue). Other AI model types don't.
3. Hallucinations based on poor quality data are likely to be worse.
comment in response to
post
Thanks. Higher education is clearly a lot of work, but if you choose a subject that's well aligned with your goals it pays off.