MIT robotics pioneer Rodney Brooks thinks people are vastly overestimating generative AI: The bona fides: βWhen Rodney Brooks talks about robotics and artificial intelligence, you should listen. Currently the Panasonic Professor of Robotics Emeritus at MIT, he also co-founded three key companies, including Rethink Robotics, iRobot and his current endeavor, Robust.ai. Brooks also ran the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) for a decade starting in 1997.β The key passage: βHe says the trouble with generative AI is that, while itβs perfectly capable of performing a certain set of tasks, it canβt do everything a human can, and humans tend to overestimate its capabilities. βWhen a human sees an AI system perform a task, they immediately generalize it to things that are similar and make an estimate of the competence of the AI system; not just the performance on that, but the competence around that,β Brooks said. βAnd theyβre usually very over-optimistic, and thatβs because they use a model of a personβs performance on a task.β > > Wonder if this is a good example of Amaraβs Law: We tend to overestimate the impact of a technology in the short run and underestimate itβs impact in the long run.
Predictions Scorecard, 2024 January 01: The author is Rodney Brooks (yes, the same guy from the first story). What I love is that not only does he make predictions but he also posts a scorecard of how well he did AND as time goes by, he includes vectors like a prediction being or becoming - accurate, too pessimistic, or too optimistic. Now I usually say (like 99.9% of the time) that forecasts beat predictions. I still believe that but if you are going to make predictions (and I believe Prof Brooks is really making forecasts) then the least you can do is include a scorecard for how well you've done over time. Has anyone done a historical analysis of the Gartner Hype Cycle and seen how closely it matches reality? Or do we just accept it as a chart of hype and not an actual prediction/forecast of the technological future? Anytime, anyone comes up with a "prediction" - shouldn't our VERY first question be "how have you done in the past?"
How to Get Root Access to Your Sleep Number Bed: I include this for no other reason than I just love the sheer geekiness of it - the just burr in the saddle of having a piece of gear in your house that you own and that you donβt have full access to. Very William Gibson.
Move over, remote jobs. CEOs say borderless talent is the future of tech work: βRemote is like the gateway drug to borderless,β said Jeremy Johnson, CEO of AI-enabled tech talent marketplace Andela, whose customers include Goldman Sachs, Github and Coursera.β And βBorderless tech hiring has doubled in the last three years, according to Gartnerβs 2023 CEO Survey. By 2022, the tech talent workforce in cities like Beijing and Delhi far outweighed that of U.S. powerhouses like San Francisco and New York, reports CBRE Global Tech Talent Guidebook 2024.β I still have my copy of Benedict Andersonβs Imagined Communities. As an anthropologist and historian, I think itβs kind of adorable when people think that the current embodiment of the nation-state in the last form in which human belonging can evolve into. I think this is a signal growing in strength that corporate leaders need to think about what their orgs will look like and what talents, skills, and roles theyβll need not just in a global world but in a borderless one.
NPR staffers share their favorite nonfiction reads of 2024: Of course Iβm going to share this list. Now I want them to do one of their fav scifi stories/books of 2024.
He Helped Invent Generative AI. Now He Wants to Save It: I applaud his efforts and he certainly has the intellectual chops but I donβt think any open source model will do what we need it to do. I donβt think any market-based solution will do what we need it to do absent any change in regulation. I think βAI safetyβ whatever that means - has to be supported by governmental regulation that has to be backed up by real penalties including the potential of jail for individuals. I also think that the development of βconsumer-side, enterprise-gradeβ AIs that can be trusted by individuals and that will fight/negotiate with other AIs on their usersβ behalf will have to become a market force. > > This is a PRIME example of where consumer-side, AI-enabled agents could work for individuals (analyzing changes to things like ToSs and advising on potential consequences) When the Terms of Service Change to Make Way for A.I. Training.
Amazon hires founders from well-funded enterprise AI startup Adept to boost tech giantβs βAGIβ team: Really hate the term βAGI Teamβ - its like being on a team for something that doesnβt exist. This to me is key - βwill accelerate our roadmap for building digital agents that can automate software workflows.β For decades, 99.9% of the work on machine learning and what we now include somewhat inaccurately as βAIβ - took place out sight of the mainstream. Thatβs one reason it took everyone by surprise - they didnβt see the heavy lifting that went on to get us to something like Chat GPT. I think this is analogous - so much of the work of AI-driven agents will take place behind the scenes and weβll just feel the effects. Thatβs a dangerous place. Consider you donβt see the work that your org is doing to create agents that streamline internal processes UNTIL the layoffs come because now they need fewer people doing the same jobs they used to do. Pay attention and learn how to deliver on higher value activities for your org. >> Orby is building AI agents for the enterprise. < < Now a lot of these companies wonβt make it but they will develop features and functionality that will and that will speed the dynamic I talk about above. And to be clear, an agent is different from an Assistant - the former is built to execute a specific task - the latter has broader applications so tougher to implement but watch the agents > > AI Work Assistants Need a Lot of Handholding.
Quoraβs Chatbot Platform Poe Allows Users to Download Paywalled Articles on Demand: This is some fine writing - βThese stupid chatbots drive me fucking crazy,β wrote Defector co-owner and editor in chief Tom Ley in a text message. βDefector does not condone having its precious blogs stolen by a dumb chatbot backed by egghead-ass Andreessen Horowitz.β
Paramount's strategy of purposeful digital rot doesn't even make sense: This is stupid and sad.
Look out, Meta Ray-Bans: These are the world's first smart glasses with GPT-4o: Gimme > > βSolos says the AirGo Vision will provide real-time information based on visual input, such as recognizing people, objects, or landmarks (similar to Meta's Ray-Ban smart glasses), or giving walking directions, checking the price of something, and more.β Another story here about Metaβs glasses and what they can and canβt do yet: I Wore Meta Ray-Bans in Montreal to Test Their AI Translation Skills. It Did Not Go Well.