danielfennelly.com
Programmer, cat dad, synthesizer knob twiddler, AI experimentalist
46 posts
87 followers
206 following
Regular Contributor
Conversation Starter
comment in response to
post
Disappointed the article doesn’t distinguish between AI systems that have data augmentation and LLMs which don’t. Not a fanboy but OAI’s mentioned hallucinations matter a lot less than Google and Bing in this context.
comment in response to
post
I’m more optimistic about the use of AI for inter-human alignment than about solving AI alignment in any general fashion. Universal alignment isn’t needed to facilitate compassionate discourse.
comment in response to
post
Making progress in post-modern ethics is far more difficult than incremental improvements to algorithms. Generative AI cranks up the volume on unsolved cultural problems.
comment in response to
post
comment in response to
post
This isn't what solidarity looks like my dude. We can call attention to the disrespect teachers face without framing it as a contest.
comment in response to
post
Does (1) imply that we've solved epistemology? There's a lot of interesting philosophy sitting right underneath the surface of our joking about GenAI.
comment in response to
post
@hankgreen.bsky.social is being funny above, but let's get serious for a moment. The problem of clearly separating bullshit from non-bullshit is an ongoing and unresolved subject of philosophy called epistemology. Hank explains more in this 10 minute video: youtu.be/kXhJ3hHK9hQ?...
comment in response to
post
Not historically. Even idiot humans possess general intelligence by virtue of adaptive behavior (intelligence) in a wide range of environments (general). We've just never had a machine (artificial) that can exhibit the same.
comment in response to
post
This is a funny joke and also a real area of computer science research (Not necessarily hot, but real nonetheless)
en.m.wikipedia.org/wiki/Ternary...
comment in response to
post
One of the features which I think isn't yet fully appreciated is how users are empowered to own their own verification. Individuals and institutions with a website and some IT capability can tie their identity to their website. blueskyweb.zendesk.com/hc/en-us/art...
comment in response to
post
I love this moment of honest question and honest answer. This is the culture I hope to see more of on this platform.
comment in response to
post
Lots of possibilities in between naive belief and zero AI. I find value in suspending disbelief and engaging AI characters, but that doesn’t mean in the larger frame I don’t know they’re mechanical. (With acknowledgment of privacy and exploitation dangers)
comment in response to
post
Hope you dig it! Looks like the artist himself is also here on BlueSky @bennjordan.bsky.social
comment in response to
post
Arboreal by The Flashbulb is one of my all time favorites.
theflashbulb.bandcamp.com/album/arboreal
IDM, ambient, jazz, post-rock, cinematic. Lush traditional instrumentation juxtaposed against sparkling digital glitching
comment in response to
post
I'm generally in favor of open source AI, but this seems straightforward. It's a ToS violation and harassment of others when they point out the ToS violation. What's so nonsense?
comment in response to
post
youtu.be/HV2UoWhV7qs?...
comment in response to
post
nuance 2: the ethics of producing a tool are connected to but distinct from the use of that tool. It can be ethical to produce dynamite and use it for construction, but unethical to bomb others. Tools may be produced which have ethical and unethical applications.
comment in response to
post
nuance 1: it is a ToS violation to publish a dataset of BlueSky data without providing a means for users to redact their data from that dataset.