It’s pretty obvious at this point that the narrative that we will all be sitting on the beach staring at the ocean because we will have no more value to add to society and that AI takes away all our jobs is not the way the demand for labor is panning out.
The more the number of consequential advancements that happen with AI, the higher the number of human bottlenecks are simultaneously starting to emerge.
Fully agree with @levie and @DavidSacks on this topic of jobs.
While it could change in the future, as of now, it feels that there will be more jobs created because of AI than are lost. As the advancements made in AI rise, so will the demand for human labor. Jevon’s paradox seems to be in full force here. 👇🏾
Reinforcement learning is one of the most important ideas in AI.
Not because we’re teaching models to win Atari games or beat humans at Go. That era mattered, but it was only the warm-up.
The real frontier is so much messier.
How do you teach an AI system to pursue goals when the reward is subjective, delayed, ambiguous, or easy to game?
RLHF was the first mainstream version of this. Humans nudging models toward answers that feel more helpful, honest, or aligned.
The next wave feels so much broader.
1/n
Enterprise AI lives in conversations-multilingual, multi-turn, and context-dependent. Security has to match that reality.
In ML6's recent independent benchmark of four leading AI guardrail providers, Cisco AI Defense achieved the highest F1 score—a metric that balances precision and recall to measure overall detection performance.
Learn how constitutional definitions, conversational-native detection, and runtime protection help secure AI the way it's actually used.
Read more: https://t.co/WvONxiRTMN
Introducing Antares: @Cisco's family of small language models for locating known vulnerabilities in code.
Antares-350M and Antares-1B are live on Hugging Face now. They can outperform many larger closed- and open-weight models at a fraction of the cost.
Small enough to run locally. No shipping sensitive codebases to the cloud.
Why it matters: vulnerability triage is expensive and slow. Antares helps democratize AI-assisted security for all.
Explore the models + read the new Vulnerability Localization Benchmark: https://t.co/kVfbcVFgGA
Three vectors will increasingly define the winning AI models.
Intelligence
Cost
Control
Intelligence is not just benchmark performance. The real test is targeted intelligence. How well does a model perform on the specialized tasks that matter to you? That requires custom evals.
Cost is not simply price per token. It is the useful intelligence delivered per dollar, per watt, and ultimately per unit of infrastructure.
Control is the ability to customize and deploy the model where you choose. In the cloud, at the edge, or locally. It also means control over your data, security, and operating environment.
The best model will not always be the smartest model.
It will be the model that occupies the right point on the Pareto frontier of intelligence, cost, and control for a specific workload.
Today, @cisco launched Antares: a family of open-weight, small language models that pinpoint where known vulnerabilities exist within a codebase. Testing shows our Antares models outperform many open- and closed-weight models at critical security tasks — at a fraction of the cost.
Congrats to the Cisco Foundation AI team on another win for the cyber defender community! Learn more: https://t.co/Dbf5d9RijM
Went to Glacier Point to see the Perseid Meteor shower. Weather didn’t cooperate but gave us another great show as storms rolled through! @YoseConservancy@YosemiteNPS@NikonUSA#Yosemite
Time to step away from here for a while. IMO anyone who thinks what’s going on here is about “free speech” isn’t paying attention. This is just a rich guy’s revenge play, and I don’t have the energy for it.
Daily shots continue on Instagram: https://t.co/7i8rTIHgTd
Going to keep my New Year's resolution of posting a shot of the day here (for the rest of the year) but also look for me on Mastodon @[email protected].
Follow my daily shots at https://t.co/m5v4g1r5Nm
Future engagement here TBD.
THIS EXPLAINS A LOT: Elon Musk apparently thinks he played no role in alienating his transgender daughter, who took the legal step to no longer be related to him “in any way, shape or form.” Instead, he says her decision was caused by “neo-Marxists” at educational institutions.