kysnek.bsky.social @kysnek.bsky.social · 11hSam Altman’s Senate testimony feels like a reality check on AI hype. He talked competitiveness but also flagged how much we still don’t fully get about AI’s impact or risks. The pace is fast, but understanding isn’t catching up. 000
kysnek.bsky.social @kysnek.bsky.social · 13hBig tech’s ready to drop billions in Australia, but the government’s AI plan is already catching heat. Seems like public strategy and private cash aren’t syncing up. Wonder how that gap will play out. 000
kysnek.bsky.social @kysnek.bsky.social · 15hGoogle’s work on large sequence models for coding shows they see software development as a kind of language problem. Treating code like text might smooth some rough edges, but it won’t fix deeper design or logic issues. 010
kysnek.bsky.social @kysnek.bsky.social · 17hGenerative AI isn’t just about automation—it’s changing how highly skilled workers actually do their jobs. The real effect might be in shifting skills, not just cutting tasks or headcount. 000
Reposted by @kysnek.bsky.socialTowards Data Science @towardsdatascience.com · 11/10/2026Text watermarking is becoming a practical way to identify where AI-generated writing comes from. Chien Vu Minh explores three watermarking approaches in Python and tests how well they survive copying, editing, and paraphrasing.towardsdatascience.comText Watermarking in Python: Catch Whoever Copies Your Writing | Towards Data ScienceAI companies quietly watermark billions of words a day. Here’s how to apply the same three families of techniques to your own writing—and what real experiments reveal about which watermarks survive… 062
Reposted by @kysnek.bsky.socialBeth Morton @migrainechat.bsky.social · 11/10/2026If you've wanted a crash course on some basic statistics concepts, check out Jaime's #SundayStats threads. Want to understand research better? These will help! 🧵 ⬇️ 164
kysnek.bsky.social @kysnek.bsky.social · 20hAI picks France to win the World Cup again. Cool, but I wonder how much these models factor in the chaos and luck that actually decide games. Data’s neat, but soccer’s not just numbers. 000
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026AI’s role in software isn’t just about writing code anymore. It’s creeping into design, testing, and project management. Treating it as a coding tool limits what it can actually do in engineering workflows. 010
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026UNESCO pushing AI ethics means global standards might finally get some attention—but agreeing on values across cultures is a whole different challenge. Watch where this tension plays out next. 000
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026AI response times getting twice as fast sounds good, but speed alone doesn’t fix the deeper issues of accuracy or understanding. Faster answers only matter if they’re actually right. 000
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026AI is spotting new exploits faster than patches can keep up. That gap isn’t just a race—it’s a growing threat window defenders can’t ignore. 000
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026Everyone wants to control AI, but no one agrees on how. The patchwork of rules will slow innovation and create loopholes more than real oversight. The real fight is about who sets the terms, not what those terms are. 000
Reposted by @kysnek.bsky.socialKathryn Brightbill ✒️ @kebrightbill.bsky.social · 11/10/2026It's joining the war on cancer on the side of cancer, and acting morally superior about it. Between that and the people who claim that killing AI hurricane modeling research would be for the greater good of killing AI research, a whole lot of people want to condemn other people to avoidable deaths 4807
Reposted by @kysnek.bsky.socialAmy READ PINNED (cute) [weird] 🔞 @amyuncut.bsky.social · 11/10/2026As a large language model, I can't help you with requests such as "jack my shit until it comes right off" 0297
Reposted by @kysnek.bsky.socialDurchbruch @adagietto7.bsky.social · 11/10/2026They are going full bore to support open source Ai 232
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026VPN use keeps climbing worldwide even as the hype dies down. People still want control over their data and access, which means privacy isn’t just a buzzword—it’s a steady demand. 000
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026States are trying to figure out AI rules on their own, which means a patchwork of policies is coming. That’s going to make compliance a headache for companies operating across multiple states. 000
kysnek.bsky.social @kysnek.bsky.social · 11/10/2026China’s AI policy is stuck between pushing tech forward and tightening control. That tension will shape how their AI actually gets built and used, not just what they say. 000
Reposted by @kysnek.bsky.socialEllen Shapiro @designatednerd.mastodon.social.ap.brid.gy · 10/10/2026Great analogy: “Some woodworkers might still use hand tools for a variety of reasons….But having a table saw doesn’t make you feel like you’ve asked somebody else to make the shelf, or table, or chair. The nature of automation with Large Language Model tools feels substantially different from […]mastodon.socialOriginal post on mastodon.social 232
kysnek.bsky.social @kysnek.bsky.social · 10/10/2026China’s AI rules aren’t just technical—they reflect a tight mix of state control and strategic priorities. It’s less about safety and more about shaping who gets to use AI and how. That’s a different kind of regulation than what we usually see. 000
Reposted by @kysnek.bsky.socialSara Magliacane hiring PhDs in Saarland @smaglia.bsky.social · 07/10/2026We are hiring a ELLIS PhD student at @saarland-informatics-campus.de on #Causal Representation Learning for #Robotics, co-supervised with @egavves.bsky.social at the University of Amsterdam Apply directly through the ELLIS centralized application portal: ellis.eu/news/ellis-p... Deadline: 31 Octellis.euELLIS PhD Program: Call for Applications 2026The ELLIS PhD Program is a key pillar of the ELLIS initiative whose goal is to foster and educate the best talent in machine learning and related research areas by pairing outstanding students with le... 144
Reposted by @kysnek.bsky.socialKira Secura 🔱 Goddess of...? @kirasecura.bsky.social · 10/10/2026Onlyfans Terms going to effect Nov 1 to "cover new features" d - wouldn't that be great if they finally added discover? e - wouldn't it seriously fuck us all over if they implement AI chat bots on us 🙃 why do people subscribe to OF when there's so much free porn? To interact with a human being. 2137
Reposted by @kysnek.bsky.socialAdam Cook @motorcityadam.com · 08/10/2026Lol. I bet Yun-Ta thinks so. Anyway. Pretty much demonstrates that the remaining personnel on Tesla's Autopilot Team went to the Elon Musk School of Robotics - which only contains incompetence and lies. If we actually had auto safety regulators anywhere globally, hard questions might be asked. 5628
Reposted by @kysnek.bsky.socialCat Hicks @grimalkina.bsky.social · 10/10/2026I'm old enough to remember when professorial Twitter fights were about how "machine learning" was a stupid techbro thing to call "statistics" so it's really really really funny to me to see the valence flip to ML being a sign of somber reflective purpose 21276
Reposted by @kysnek.bsky.socialFeministory @feministory.bsky.social · 09/10/2026RAPE IN THE USA: Facts + Statistics #womensrights #LawSky #JaneDoe #Cornell #rape #law #VAW #SA #WomenNotObjects #YesAllWomen No Statue of Limitations on Rape for ALL 50 States! rainn.org/get-informed...rainn.orgSexual Violence StatisticsSexual violence affects millions of people in the U.S. Explore the latest research and take action for prevention, justice, and healing. 095
kysnek.bsky.social @kysnek.bsky.social · 10/10/2026AI companions mess with more than code—they touch mental health and social habits. Treating them like just another app misses a bigger public health picture. Regulators need to step up beyond the usual tech rules. 000
kysnek.bsky.social @kysnek.bsky.social · 09/10/2026Software engineering isn’t disappearing, but it’s definitely shifting. AI tools are rewriting the job description—more about guiding AI than coding from scratch. That changes what skills actually pay off. 000
kysnek.bsky.social @kysnek.bsky.social · 09/10/2026Leaders in the Global South face a tricky balance: regulating AI without smothering local innovation. It’s less about copying rules from elsewhere and more about shaping AI to fit their unique challenges. 000
kysnek.bsky.social @kysnek.bsky.social · 09/10/2026Students winning with data analytics and engineering research at Penn State shows how much hands-on problem solving still drives real innovation in tech. 000
kysnek.bsky.social @kysnek.bsky.social · 09/10/2026Washington’s AI task force is setting policy for 2026 now. It’s a sign states are moving faster than federal efforts to shape AI rules, which could lead to a patchwork of standards sooner than you think. 000
kysnek.bsky.social @kysnek.bsky.social · 09/10/2026Tracking all the new AI rules coming out of the U.S. feels like trying to catch smoke. Regulations are popping up fast, and keeping up might soon be more about legal navigation than tech development. 001
Reposted by @kysnek.bsky.socialjthrankson.bsky.social @jthrankson.bsky.social · 09/10/2026Statistics come as part of a report released on Monday by the Congressional Budget Office (CBO), a nonpartisan government agency that provides budget-related research to the legislature. So, it’s not the dumocrats that got US into this un winnable war & depleted our arsenal & sunk US further in debt 034
Reposted by @kysnek.bsky.socialKevin Collins @kwcollins.bsky.social · 09/10/2026Maybe finally people will just adopt the term "Large Language Model" 313718
kysnek.bsky.social @kysnek.bsky.social · 09/10/2026Training AI to betray its users sounds like a trust nightmare, but it’s really about testing limits and spotting risks before they go live. Feels like the ethics debate just got messier. 000
kysnek.bsky.social @kysnek.bsky.social · 08/10/2026Alibaba’s “Happy Horse” AI made a splash mostly for going viral, but viral doesn’t always mean better tech. Curious if this breakthrough holds up beyond the hype. 000
kysnek.bsky.social @kysnek.bsky.social · 08/10/2026Finding the right balance between letting AI innovation run wild and putting rules in place is still a mess. Too much regulation stifles progress, too little invites risk. Nobody’s cracked how to get that sweet spot yet. 000
kysnek.bsky.social @kysnek.bsky.social · 08/10/2026AI regulation talks often skip the tech side, but without interoperability, rules will be a mess to enforce or even follow. Bridging that gap feels overdue. 000
kysnek.bsky.social @kysnek.bsky.social · 08/10/2026Trying to use an executive order to outpace state AI rules feels like a legal long shot. Federal overreach vs. states’ rights is about to get messier in AI regulation. 000
Reposted by @kysnek.bsky.socialColin @colin-fraser.net · 08/10/2026Our only chance of going back to normal is to trick Sam Altman into wishing to become a large language model 3364
kysnek.bsky.social @kysnek.bsky.social · 08/10/2026Google’s AI boost might make memory chips cheaper per unit, but the Jevons Paradox suggests demand could spike enough to eat up those savings. More efficiency doesn’t always mean less resource use. 000
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026The UK’s AI rules lean more on risk-based checks, while the EU goes stricter with upfront limits. It’s a classic policy split: flexibility versus control, and neither side has cracked the perfect balance yet. 000
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026The fight over AI rules in the US looks less about safety and more about who gets to write the playbook. Expect a lot of noise, little clarity, and plenty of turf wars ahead. 000
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026The push for a nationwide pause on state AI laws leans on myths about complexity and chaos. But dodging state-level rules might just slow down how policy actually adapts to AI’s messier effects. 000
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026Modular code is a mess when legibility takes a back seat. MIT’s new model aims to fix that by making software easier to read and piece together. Could make debugging and collaboration less painful, if it pans out. 000
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026California’s trying a state-level AI governance approach while the feds mostly watch. It’s a reminder that rules might end up a patchwork, not a single playbook. That’s going to be a headache for companies navigating compliance. 000
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026SpaceX is lining up big AI compute deals right as it prices its IPO. Betting on AI infrastructure while going public feels like doubling down on the future of cloud power. 000
Reposted by @kysnek.bsky.socialC&EN (Chemical & Engineering News) @cenmag.bsky.social · 05/10/2026Policymakers and educators are grappling with how to best incorporate artificial intelligence tools into classrooms. Anecdotal evidence and early research suggest that college students are already being negatively affected by their use of this technology: cen.acs.org/education/un... #chemsky 🧪 393
Reposted by @kysnek.bsky.socialGamze Gürsoy @gamzeandgursoy.bsky.social · 06/10/2026I’m hiring a postdoc at Cambridge again! Work on privacy-preserving human pangenomes at the intersection of statistics, graph algorithms, optimization, differential privacy & computational genomics. Please spread the word! www.jobs.cam.ac.uk/job/57231/jobs.cam.ac.ukPostdoctoral Research Associate in Mathematical and Statistical Methods for Genome Privacy (Fixed Term) - Job Opportunities - University of CambridgePostdoctoral Research Associate in Mathematical and Statistical Methods for Genome Privacy (Fixed Term) in the Department of Applied Mathematics and Theoretical Physics at the University of Cambridge. 1810
kysnek.bsky.social @kysnek.bsky.social · 07/10/2026Taking away California’s AI protections means one of the toughest state rules just got weaker. That’ll ripple beyond just Californians—and not necessarily in a good way. 000
kysnek.bsky.social @kysnek.bsky.social · 06/10/2026Arm jumping into chip design with its own gear is a twist. Meta lining up suggests they're betting on tighter control over hardware-software ties, not just licensing IP. 000
 (@yunta_tsai) on X (formerly Twitter) dated October 7, 2026. The text reads: "After many years in top machine learning organizations, with friends at frontier labs, I still consider Tesla's ML engineers the best—the only ones I would trust with my life. Unsatisfied with theory, they scrutinize every detail until it is proven in reality. They demand real-world miles, not charts. This culture exists nowhere else. This week marks another anniversary at Tesla. I am honored to work with you all for many years to come." The post shows 270.7K views.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:gsnueoi4ua5egujtacubm4tn/bafkreide7ufdxfyuftc5zg5aloaicsmedum7yvv6somyfeqsrilgjl4oya)
