Dr. Rachael Tatman @rctatman@mastodon.rctatman.com
Developer and data science educator who specializes in language technology. Making NLP boring. Linguistics PhD. Data science, NLP, Stats, ML, R, Python, FAccT. Tweets my own & CC-BY-SA. She/her.
Joined Apr 2022
Here's the link for our weekly coffee chat tomorrow! It'll probably a bit more research-heavy for the next couple weeks with NeurIPS & EMNLP coming up. ☕️😁
https://www.youtube.com/watch?v=_T-Y0k8TEV0
This is (unfortunately) a very good example of how paternalism and ableism in language technology not only directly harms Deaf folks (by spreading false information) but also results in a lot of wasted money & effort building genuinely useless systems.
RT @frantciscoch@twitter.com
This is VERY offensive and false
🐦🔗: https://twitter.com/frantciscoch/status/1574902862299254784
Goood morning! Stream starts in about half an hour, see you there! (Or on the replay if you miss it. 😉)
https://www.youtube.com/watch?v=mj--nbUopHo
RT @sarahookr@twitter.com
This is a really important reminder. I highly recommend to those applying -- get someone else to look over your personal statement + look up resources on how to write a compelling personal statement before getting started. You don't have much space, so use it meaningfully. 💙 https://twitter.com/forai_ml/status/1574503707940642837
🐦🔗: https://twitter.com/sarahookr/status/1574528714804809743
Virginia folks: if you've got a minute to stand in solidarity with trans kids I'd appreciate it if you left a comment to *oppose* the rollback of protections for trans students in schools.
RT @SallyLHudson@twitter.com
Public comment on the rollback of Virginia’s model policies for trans students is now open. You can make your voices heard! 🏳️⚧️
https://townhall.virginia.gov/L/comments.cfm?GDocForumID=1953
🐦🔗: https://twitter.com/SallyLHudson/status/1574374294443941889
For my stream tomorrow I'll be doing something I've gotten some requests for: ranking different ML/NLP companies by how willing I'd be to work there.
(Recruiters: I am NOT on the market, please don't bother.😅)
https://www.youtube.com/watch?v=mj--nbUopHo
RT @signalapp@twitter.com
For people connecting to Signal in Iran, here's more information on how to use a proxy, in Persian. Please share: https://support.signal.org/hc/fa/articles/360056052052-%D9%BE%D8%B4%D8%AA%DB%8C%D8%A8%D8%A7%D9%86%DB%8C-%D8%A7%D8%B2-%D9%BE%D8%B1%D9%88%DA%A9%D8%B3%DB%8C
#IRanASignalProxy
🐦🔗: https://twitter.com/signalapp/status/1573347802314416132
RT @tracykteal@twitter.com
We're hiring @rstudio@twitter.com for roles in developer relations!
📗 Interested in coding, education & advocacy?
🚀 Come work with us & help users learn and flourish!
* Developer Educator - #tidyverse package development
* Developer Advocate - Quarto #quartopub
https://www.tidyverse.org/blog/2022/09/devrel-hiring-2022/
🐦🔗: https://twitter.com/tracykteal/status/1573347761315258368
Language modelling as a task was originally developed as part of a larger pipeline and (fun low-stakes art projects excluded) I'm yet to be convinced that just making the models way bigger changes that. 🤷♀️
As we learn more about adversarial attacks on LLMs, the best overall mitigation strategy continues to be simply *not directly serving LLM outputs to users*.
It is wild to me how much of generative ML boils down to: "We would rather spend millions on dubious data collection methods and training shiny new models than actually pay any artist to do their art." (Especially with the assumption that generated art would be just as good.)
RT @harrietprebble@twitter.com
1/7 There's a lot of talk in the games industry at the moment about the use of AI technology for voice work. As someone who works for a game developer and is also a professional voice actor, I have some thoughts. A thread. 🧵↓
🐦🔗: https://twitter.com/harrietprebble/status/1573084171719020544
Good morning friends! Our coffee chat/ML & NLP week in review roundup starts in just about ten minutes if you want to grab a coffee/water/snack first.
This is utterly indefensible.
RT @WolfieChristl@twitter.com
According to contracts and a whistleblower, a private cybersecurity firm is selling massive amounts of global internet traffic data to the US military, from netflow to package capture data, including on 'URLs accessed', 'cookie usage', email/PCAP and more:
https://www.vice.com/en/article/y3pnkw/us-military-bought-mass-monitoring-augury-team-cymru-browsing-email-data
🐦🔗: https://twitter.com/WolfieChristl/status/1572682780781481985
Here's the link to tomorrow's coffee chat stream! It's shaping up to be a little more research-focused than usual this week.
(Also I'm getting a COVID booster today so there's a sliiight chance this might be rescheduled.)
https://www.youtube.com/watch?v=Rgmr9kf5WbA
Would it be weird to put that I'm a Kaggle grand master in my Twitter bio? I don't personally put that much stock in it but it might reduce the amount of ml-splaining I get. 🤔
My stream on figuring out roughly how much data you need for a NLP project--and how to adjust your project depending on how much data you actually have--is starting in about half an hour.
See you there or on the replay!
Here's the link for tomorrow's stream! We'll be talking about how much data you *actually* need for a given NLP project & some guidelines to help you figure that out before you start.
RT @RVA_Votes@twitter.com
If I were you, I wouldn’t want to be 1 of those 4 people 😳 Don’t forget to register to vote! Deadline is October 17th!! #VoteReady https://twitter.com/natlvoterregday/status/1570183175195824130
🐦🔗: https://twitter.com/RVA_Votes/status/1571868058800496643
RT @PennDDDI@twitter.com
Learn good practices for data organization e.g. file naming and documenting research, @upennlib@twitter.com data management workshop on 9/20! Register at https://libcal.library.upenn.edu/event/9460188?hs=a
Developer and data science educator who specializes in language technology. Making NLP boring. Linguistics PhD. Data science, NLP, Stats, ML, R, Python, FAccT. Tweets my own & CC-BY-SA. She/her.
Joined Apr 2022
