The Distributed AI Research Institute (DAIR)

🔇🔇🔇 This Monday September 9 at 8am PT/5pm CET will be the next panel in our Data Workers’ Inquiry series, where we compare the type of training data workers need vs what they receive instead. ➡ Register at https://lnkd.in/eAZTeYJ8. Featuring ➡ Fasica Berhane who was a journalist fleeing #TigrayGenocide only to be exploited as a content moderator in Nairobi and then get laid off for attempting to form a data workers’ union. Read her report on the toll of experiencing the genocide AND moderating that content for Meta during it, including mental health issues and drug abuse experienced by content moderators: https://lnkd.in/eMgu_pNS. ➡ Yasser Alrayes works as a data annotator while studying computer science. He is passionate about technology and eager to apply a blend of academic knowledge and project experience in the tech industry. His vlog highlights the inadequate training of data workers in Syria and their resulting personal struggles. It advocates for fair, structured, and honest training processes to empower and prepare workers for the job. Watch the video at https://lnkd.in/e-J8FFgd. TechCrunch says it is "well worth eight minutes of your time"! ➡And our very own Adrienne Williams, and Dr. Milagros Miceli will discuss how and why data workers are trained to "think like machines" and how each one of us is being used as a "zombie trainer”, a term coined by Adrienne in her Newsweek op-ed you can read at https://lnkd.in/eduNrWSW. ➡ Find the future talks in the series at data-workers.org/events and register now for the ones you want to attend. The repository of reports, animation shorts, essays, vlogs, podcasts and more produced by data workers is at data-workers.org.

There are some real NLP tools that save people time, and for which there could be tools in many other languages. -Spell checkers. -Relatively good auto generated captions that you check and fix after the fact. -Speech to text tools that help people (e.g. journalists) transcribe their recordings, or where you can search for specific parts of talks that you remember something was said but don't remember where, etc. There are real tools and then there is AI Hype.

A chilling anecdote from one of the individuals contributing to the Data Workers' Inquiry. Check out our podcast for the full conversation! #podcast #ai #dataworkers #labor https://lnkd.in/dJFbRRh4

What kind of job training did you get when you first started out working? What kind of training do you think #DataWorkers get, who shape #AI and any online platform you can think of? Join us to discuss this next Monday with the Data Workers' Inquiry! 🗓️ September 9, 5 pm - online https://lnkd.in/enS_P_TG Panelists: Fasica Gebrekidan, Yasser Alrayes, Adrienne Williams, Dr. Milagros Miceli Moderation: Laurenz S. Let's not forget: behind the glossy veneer of #AI lies a hidden workforce of data workers who form the backbone of this technological revolution. And while we can debate about whether the #hype is over or not - data workers are not going anywhere. The workforce will be growing, and we need to fight for working conditions to improve. The Team: Dr. Milagros Miceli, Adio-Adet Dinika, Ph.D, Krystal Kauffman, Camilla Salim Wagner, Laurenz S. Support from: The Distributed AI Research Institute (DAIR), Timnit Gebru, Turkopticon, SUPERRR Lab, Writers Guild Of America, USA, Guerrilla Media Collective, Arte es Ética

Congratulations to Robert Williams, the winner of AJL’s first Gender Shades Justice Award, for finally reaching a settlement agreement in his lawsuit with the ACLU of Michigan against the Detroit Police Department. Williams was arrested for shoplifting based on faulty facial recognition technology, and since then, he has been working tirelessly to protect others from the dangerous uses of this technology. You can read more about Robert Williams and his case in his TIME op-ed: https://shorturl.at/L8xK9 and more about AJL’s Gender Shades Justice Award in this press release: https://shorturl.at/jIJ3N

In recognition of International Translation Day, ELP invites you to join us on September 30 at 10:00 am Pacific time for a free, online public forum: “Building community networks of dignity and solidarity: Indigenous language interpretation is a human right.” https://lnkd.in/guVZC-QD   Indigenous migrants face discrimination and violations of their linguistic and human rights every day. This public forum will highlight the voices of four grassroots Indigenous-led organizations doing the work of Indigenous language interpretation in communities around the US in this critical moment.   ELP’s goal in offering this forum is to provide a space for the public to have the chance to listen to those on the ground doing this work - and to create the possibility for public engagement, education, sensitization, and solidarity. The audience will have the opportunity to engage with the organizers at the end of the forum.   This event is made possible in part by the generous support of the First Peoples’ Cultural Foundation (http://fpcf.ca).   The panel will feature:   Ernesto Ajanel and Juanita Cabrera Lopez, International Mayan League (https://lnkd.in/ekrDhszr) Arcenio Lopez, Mixteco Indígena Community Organizing Project (http://mixteco.org) Saul Rivera, Concejo de los Pueblos Originarios (https://lnkd.in/gZpY2vya) Puma Tzoc, Pueblo Unido PDX (https://lnkd.in/gVKpUTnr)   We invite you to join us as we stand in solidarity to support the critical work that these community organizations are doing every day. Please spread the word, and we hope to see you there!   Register at https://lnkd.in/guVZC-QD

Don't miss this online panel on Monday if you're curious about how AI is truly built! Fasica Berhane, Yasser Alrayes, Adrienne Williams, and I will dive into how and why data workers are trained to "think like machines" and how each one of us is being used as a "zombie trainers." Curious? 📅 Register here: https://lnkd.in/dBnyEdmj ✅ Check out the Data Workers' Inquiry repository and future events: https://data-workers.org/ 🙏 And help us spread the word!

Heartbreaking text on the International Day of the Dissapeared, touching on hopeful imaginations of putting technology at use for families of refugees - who often don’t know what happened to their loved ones, and live in a permanent state of limbo.

We are proud to support Logic Magazine's Palestinian Journalist Fellowship in Collaboration with Arab Reform Initiative. "The purpose of this program is to support community-led storytelling on Palestine and its relationship to technology, to strengthen the magazine’s contributions to Palestine reporting, and to redistribute resources to Palestinian journalists." Read more👇 https://lnkd.in/d27kNbWD

🪲 Excited to announce the publication of our latest work InkubaLM! We're thrilled to introduce InkubaLM, a powerful yet efficient small language model (SLM) with just 0.4 billion parameters. Despite its size, InkubaLM achieves performance comparable to much larger models trained on more extensive data. 🔑 Key Highlights: - 🛠️ Versatile Performance: Excels in tasks such as machine translation, question-answering, AfriMMLU, and AfriXNLI. - 🌍 Linguistic Coverage: Supports a range of languages, including Hausa, Yoruba, Swahili, Xhosa, and Zulu. - 📉 Efficient Training: Trained on only 2.4 billion tokens. - 📚 New Datasets: Introduces two datasets—Inkuba-Mono & Inkuba-Instruct. - 💪 Inspired by Nature: Named after the mighty dung beetle, known for its incredible strength. 👉 Explore the model on Hugging Face: https://lnkd.in/dSRfbkaE 👉 Read the full paper on arXiv: https://lnkd.in/dtvuBBtS 🌍 Join us in pushing the boundaries of language modeling and bringing more representation to African languages!