
Travis Hoppe
@metasemantic
Followers
2K
Following
12K
Media
382
Statuses
2K
post-punk scientist / previous @DCHackAndTell organizer / creator of many, many irrelevant things / fan of sedenions
Washington, DC
Joined June 2015
Great panel today with Stancy Dean, @CMSGov and @ACFHHS! Bringing tech policy, AI, and algorithms to those on the ground making change at #APHSASummit2024.
Deputy Undersecretary Stacy Dean joined a great panel on the federal gov’t’s approach for AI in human services--opportunities and risks @APHSA1's Human Services Summit. Great to collaborate with @ACFHHS, @CMSGov and @WHOSTP! #APHSASummit2024
0
0
3
RT @nickfloats: You can generate some pretty incredible 360º photos in Midjourney using the --tile parameter. Once upscaled in MJ & magnifi….
0
246
0
RT @gojomo: what 'transorthogonal linguistics' (by @metasemantic) did for individual words….
0
2
0
RT @lmsysorg: 🔥Excited to introduce LMSYS-Chat-1M, a large-scale dataset of 1M real-world conversations with 25 cutting-edge LLMs!. This da….
huggingface.co
0
87
0
RT @BarbaraLancho: "We operationally define the term “dark citation” as any citation that does not include an indexed representation.". Int….
0
3
0
Our new paper "Dark citations to Federal resources and their contribution to the public health literature", Keralis et. al., looks at all the URLs in PubMed references that point to US Federal government websites.
frontiersin.org
The term "dark citations," which has been previously used to refer to citations of information products outside of traditional peer-reviewed journal articles...
1
1
1
Ever wonder how many "dark citations" there are? These are citations that don't point to a paper/book but just a website? We tracked them all! 1/N.#pubmed #AcademicTwitter
2
2
2
With all the wonderful all-purpose LLMs these days, sometimes a small fine-tuned model works best! At @NCHStats we released a model to detect non-response from Internet surveys. Basically, a scanner for gibberish, refusals, & uncertainty! . On @huggingface.
huggingface.co
0
1
1
Here's another fictional one about rats that are sad that they don't get treats anymore. Notice how #chatgpt added words like reinforcement, negative emotional response, and animal welfare:
0
0
4
Holy shit. Not only can #chatgpt dumb-down text "like a 5-year old" it can revise a poorly written abstract into scientific jargon, "I'm going to give you a basic scientific abstract and I'd like you to rewrite it so it was ready for a top-tier journal"
4
3
36
DC friends! I'm exhibiting tonight at the RAW CATALYST show held at @CityWineryDC #AIArtwork #ai #ArtificialIntelligence . Doors open at 7
0
1
3
RT @pinar_demirdag: Friends, if you are in DC, check out the exhibition of @metasemantic He has been doing marvels with AI and font design,….
0
2
0
Directly running input from GPT-3 as an eval is such chaotic energy. The thread even indicates that it can delete your hard drive if you ask it to. Very Suicide Linux OS vibes.
"You are GPT‑3, and you can't do math": Prompting GPT‑3 via zero-shot instruction to answer calculation/math questions by consulting a Python REPL.
0
0
7
RT @Chase_Davis_: 💡neural compression for textures and meshes is going to make web 3D so much more accessible. Would not at all be surprise….
0
3
0
I've already made a bunch of textures with it, but I'm sure there are plenty more to discover! Would be really interested if someone could hack around with it and figure out a different symmetry other than basic rectangular tiling.
Just figured out something really fantastic with #stablediffusion, by slicing and clever masking you can make an infinitely tiled image!
7
1
43
CC'ing those interested: .@Attic_Wizard @Aeium @arnicas @andrewwhite01.@ToolTrackers @EMostaque @alexjc @Norod78.
1
0
10