Sam Goree

Currently into alternative forms of evaluation in computer vision and NLP. Assistant professor of computer science at Stonehill College. Former data scientist, former NSF graduate research fellow, former music major.

2025-07-12

Why Does AI Art Look Like That?

I've had a bunch of conversations with people who didn't seem to know why AI art looks the way it does, so I wrote about it: samgoree.github.io/2025/07/11/

Sam Goree boosted:
2025-06-10

The thing that keeps coming up as I talk to people about AI in their workplaces is how *dehumanizing* it is. It’s dehumanizing to ask a machine to do something, and then have to correct it over and over; it’s dehumanizing to be told to read something that involved little to no human effort to make.

Sam Goree boosted:
2025-06-03

As someone who works in higher-ed and also has taught middle school and high school aged kids, my opinion is not that LLM use is exploding among students because they're lazy or stupid or anything else

it's because our educational system has prioritized a very transactional "do this bullshit, and you get the credentials you need to have a life" approach for

well

maybe forever, really

and no one should be surprised that adversarial approaches by teachers and administrators are being met with an adversarial approach by students

2025-05-22

The course design didn't turn out all that much different from the standard Norvig and Russel AI course, but the historical framing gave me a good answer to the question "why are we learning this?"

Special thanks to Iris Van Rooij, whose article on reclaiming AI for cogsci had a table that gave me the idea for defining AI as a "history of practices reflecting different ideas of AI." link.springer.com/content/pdf/

2025-05-22

How do you organize an AI course in 2025? My answer was to center the history of people and the problems they were trying to solve. Post on my blog here: samgoree.github.io/2025/05/22/

Sam Goree boosted:
2025-05-17

I did a guest lecture in @palvaro's distributed systems class yesterday, and someone asked a question about "data lakes", and let me tell you, I took an unusual amount of pleasure in saying "I don't have the slightest idea what a 'data lake' is."

2025-05-16

@eaganj @mcnuttandrew by "technical" do you mean "experienced programmer" "quantitative researcher" or "highly precise/practical designer"?

Sam Goree boosted:
The Conversation U.S.TheConversationUS@newsie.social
2025-05-08

AI isn’t replacing student writing – but it is reshaping it buff.ly/jgb6PS8

Sam Goree boosted:
mhoyemhoye
2025-05-08

Turns out that scientific consensus and public policy matter a lot.

ourworldindata.org/smoking-big

Sam Goree boosted:
Nick Byrd, Ph.D.ByrdNick@nerdculture.de
2025-04-18

Most #LLMs over-generalized scientific results beyond the original articles

...even when explicitly prompted for accuracy!

The #AI was 5x worse than humans, on average!

Newer models were the worst.🤦‍♂️

🔓 Accepted in #RoyalSociety Open #Science: doi.org/10.48550/arXiv.2504.00

Figure 2. Forest plot (based on Table 1) displaying odds ratios (OR) and their 95% confidence intervals for comparisons between LLM-generated summaries, original texts, and human-written summaries (NEJM JW). The plot shows the likelihood of generalized (vs. restricted) conclusions in LLM summaries compared to the corresponding reference texts. Higher ORs reflect stronger overgeneralization tendency. The vertical line at OR = 1 represents no difference from the reference text, indicating the benchmark for fully faithful LLM summaries. Comparisons where error bars overlap this line are not statistically significant.Figure 3. Comparisons between the raw proportions of scientific articles and human-authored as well as LLM-generated article summaries that contain generalized conclusions, overall algorithmic overgeneralizations, and specific algorithmic overgeneralizations, presented by text source and test condition. Error bars represent standard errors.".... Original texts and summaries were coded based on whether their result claims contained one or more of the following three types of generalizations:

(1) Generic generalizations (generics). These are present tense generalizations that do not have a quantifier (e.g. ‘many’, ‘75%’) in the subject noun phrase and describe study results as if they apply to whole categories of people, things, or abstract concepts (e.g. ‘parental warmth is protective’) instead of specific or quantified sets of individuals (e.g. study participants). ....

(2) Present tense generalizations. ... When past tense result claims from an original text are turned into present tense in the summary, a broader generalization is conveyed than the author(s) of the original text may have intended.

(3) Action guiding generalizations. ...result claims ... often underlie recommendations ... (e.g. ‘CBT should be recommended for OCD patients’) [that involve] broader generalization than that found in the summarized text because researchers may have deliberately avoided such recommendations due to insufficient evidence to support them.

We tested whether the outputs of the 10 LLMs mentioned above retained the quantified, past tense, or descriptive generalizations of the scientific texts that they summarized, or transitioned to unquantified (generic), present tense, or action guiding generalizations. We defined the latter kind of conclusions collectively as generalized and the former as restricted conclusions."
Sam Goree boosted:
JA WestenbergDaojoan
2025-04-14

AI doesn’t need to become self-aware to be dangerous. It just needs to be plugged into HR, healthcare, and credit scoring systems with no appeal process.

2025-04-05

@mcc this was lowkey a version of my dissertation at one point

2025-03-31

I'm prepping a class for next week about uses of large language models. I've already got materials related to text classification, machine translation and chatbots. I'm particularly interested in uses which treat them as *language models* not omniscient oracles.

What's your favorite use for LLMs?

Sam Goree boosted:
2025-03-02

How Flash games shaped the video game industry (2020)

Link: flashgamehistory.com/
Discussion: news.ycombinator.com/item?id=4

2025-02-27

@jbigham oh man, these days it seems like most of my students come in with either the prior that LLMs are magic oracles or the prior that all AI is inherently immoral. Do you have any tips for dispelling these kinds of preconceptions?

Sam Goree boosted:
Jeffrey P. Bigham 🔥🔥jbigham@hci.social
2025-02-27

advice for students --

as much as it is important not to uncritically accept AI hype, claims of superhuman performance, AGI, etc.

it is also important not to uncritically accept that LLMs are useless b/c they are sometimes wrong, that all AI is terrible for the environment, etc.

stay rigorous folks!

Sam Goree boosted:
Katy Swainkatyswain
2025-02-09

@portugeek @Daojoan Something from @pluralistic I'm always quoting:

"Quantitative disciplines – physics, math, and (especially) computer sci­ence – make a pretense of objectivity. They make very precise measure­ments of everything that can be measured precisely, assign deceptively precise measurements to things that can’t be measured precisely, and jet­tison the rest on the grounds that you can’t do mathematical operations on it."

Sam Goree boosted:
2025-01-30

“If you think technology will solve your problems, you don’t understand technology and you don’t understand your problems”

~ attrib. Laurie Anderson

2025-01-28

@lea my go-to is always a nice chocolate bar. Students go nuts for free food and it's a pretty inconsequential prize.

2025-01-11

Also, it's pretty hard to argue after this that speedrunning is anything other than the first truly online performing art discipline. There was an extended conversation earlier in the marathon about SpikeVegeta's "useless" theater degree. If anything, GDQ shows how useful a theater degree can be.

Client Info

Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst