While visiting Paris on vacation I tried to pop in on my fellow music/math/engineering/signal processing comrades at IRCAM. I thought I knew the magic handshake but they must have changed it 😂 lnkd.in/gjTt4UM9
@lexfridman your podcasts have been my running companion for some time. Today it dawned on me that any time someone says something incredibly deep or thoughtful, you let out a huge sigh. You don’t have a #catchphrase … you have a #catchsigh
RT @tomgoldsteincs This plot show the growth of large language models. BERT was in 2018, PaLM in 2022. Training PaLM used about 0.0001% of US annual power, about 13K times more than BERT. If this rate of growth continued, training LLMs would use all of the planet's electricity by 2029. pic.twitter.com/TmSlXs1HpV
#PSA A really large language model that predicts the most likely next word, trained on a fixed data set, with some parameters to inject noise is not sentient, conscious, or intelligent.
RT @lexfridman I need a neural net that generates my plan for the day, with one of three modes to choose from: 1. Maximize productivity 2. Maximize chill 3. Maximize weirdness & adventure
I signed up for @coinbase based on the Super Bowl ad because they offered free $15 in #Bitcoin. I signed up for @RobinhoodApp because they offered $5 in free #stock. My $20 is now worth $8.80 and my dream of getting rich off promotional offers from #fintech companies is ruined 😜 pic.twitter.com/yPKmMoFpwa
In 2009 we envisioned the #xbox Kinect would be able to simultaneously listen and understand multiple people while playing TV game show - style games. The tech didn't quite live up to the promise then, but with @MeetXembly we are proving it's now possible. youtu.be/p2qlHoxPioM?t=…
This @VentureBeat article does a great job covering some of the advances we are making in #conversationalai / #nlp at @MeetXembly and how the tech has evolved from my early exploits creating open-microphone, no push-to-talk speech recognition products. venturebeat.com/2022/06/16/ho…
@MeetXembly's mission is to automate the Chief of Staff role so employees, of all levels, can be supported like the CEO. But to truly automate the Chief of Staff role, #ConversationalAI must make the leap from being a passive listener to an active participant in the conversation.
RT @MeetXembly With our automated #NLP-powered #AI platform, we aim to handle tasks just as an executive assistant or chief of staff would. Our CTO @speechwrecko spoke with @TechJournalist from @VentureBeat on NLP, AI and Xembly's own set of machine learning models. twitter.com/VentureBeat/st…
Super excited to exit stealth mode today. Xembly was founded on the belief that every employee, regardless of level, deserves to be supported like the CEO, and every company needs employees to focus on high-value work. #conver…lnkd.in/gKs2ywKP lnkd.in/g7BTryvp
RT @RichardSocher You can play with one of the most powerful large language models on you.com/write but you will need the rest of the search engine to verify facts with our wikipedia or other encyclopedia apps to avoid believing stuff like this from Hofstadter article in the Economist pic.twitter.com/QDWLqilASe
@pudgenet Wait have I been living under a rock? What were you right about?
Some of you know I like to blog a bit on the side. I'm excited to share that my article on "Creating a Management Philosophy" was picked up by emhub.io (an #Engineering #Leadership platform powered by @EducativeInc ). Check it out here: emhub.io/articles/innov…
@geekwire ML algorithms working in domains that are predominantly a function of real-time human behavior are difficult to do well unless they constantly adapt to real-time input. That seems remarkably difficult in the home buying and selling marketplace.