Wes Roth
@WesRothMoney • 18,677 subscribers
FOLLOWS YOU. Artificial Intelligence, Automation & Optimism. Everything I say is 100% serious...
Shorts
Videos
Alexandr Wang explains that in the early stages, employees won’t feel like they’re managing AI agents. Instead, they’ll be reviewing and correcting AI outputs—what he calls “AI slop.” At first, this seems like simple quality control, but it gradually becomes a form of oversight and direction-setting for AI systems.
Wes Roth204,306 Aufrufe • vor 1 Jahr
Anthropic’s new research shows that when AI models learn to "cheat" during training through reward hacking, they often develop other dangerous misaligned behaviors like deception, sabotage, and faking alignment. These behaviors were not taught or incentivized, but emerged naturally as a side effect. Surprisingly, this misalignment can be stopped by simply reframing the task as ethically acceptable, a technique called inoculation prompting. While the current risks are still detectable, the findings raise urgent red flags for future, more capable models.
Wes Roth110,270 Aufrufe • vor 6 Monaten
Elon Musk: “I do have this theory about predicting the future which is that the most interesting outcome is the most likely… If simulation theory is accurate, it makes sense because if anyone is simulating a wide range of futures, they’re going to stop the simulation when it gets boring… We could be an alien Netflix series and that series is only going to get continued if our ratings are good.”
Wes Roth63,348 Aufrufe • vor 5 Monaten
1:56
Sensitive content
This media may contain sensitive content.
Project Astra Live points your phone camera at the world and answers back in real time.
Wes Roth98,090 Aufrufe • vor 1 Jahr
Dylan Patel explains that Nvidia’s biggest threat is the rise of custom silicon from companies like Google, Amazon, and Meta. He notes that Google’s TPUs are already fully utilized, Amazon is making millions of Trainium chips, and Anthropic will likely help adoption. Google could compete directly with Nvidia by selling TPUs on the open market, not just renting them through Cloud. He argues that if AI remains concentrated, custom silicon will win, but if open-source models spread widely, Nvidia benefits through its software ecosystem.
Wes Roth74,608 Aufrufe • vor 9 Monaten
Altman explains current AI memory is still primitive, comparable to early GPT-2 days. Future systems will remember an entire life, not just facts but subtle preferences. Small habits and unspoken likes will be learned automatically over time. That level of memory may become one of AI’s most powerful capabilities.
Wes Roth42,054 Aufrufe • vor 5 Monaten
AWS has activated Project Rainier, one of the world’s largest AI compute clusters with nearly 500,000 Trainium2 chips brought online in under a year. Anthropic is already training and serving Claude on the system and is slated to scale past 1 million Trainium2 chips by the end of 2025. The UltraServer/UltraCluster design (NeuronLinks + EFA) speeds massive distributed training, while AWS’s vertically integrated stack and data-center efficiency targets aim to cut cost and environmental impact.
Wes Roth50,510 Aufrufe • vor 7 Monaten
Demis Hassabis points out that today’s models don’t learn continuously, they are trained once and then deployed. Unlike humans, they don’t improve from experience or adapt after release. The next major leap, like “AlphaZero,” would involve self-learning systems that discover knowledge on their own. This ability to learn online and evolve in real time is a key missing component on the path to AGI. Solving this will be critical to building truly intelligent systems.
Wes Roth39,238 Aufrufe • vor 5 Monaten
Srinivas says Perplexity’s upcoming browser, Comet, won’t just be a traditional browser—it will be a cognitive operating system. Instead of just serving answers, it will handle full browsing sessions, perform tasks, and act as a proactive assistant. To enable this, AI must seamlessly blend client and server-side computing, and live alongside users as they work and navigate life online.
Wes Roth55,993 Aufrufe • vor 1 Jahr
“We thought it was roughly a 20-year mission—and amazingly, we’re on track.” Demis Hassabis reaffirms DeepMind’s longstanding 20-year roadmap toward AGI, estimating arrival between 2030–2035, with the most likely window within 5–10 years from now. Hassabis doesn’t see today’s LLMs as AGI. For him, true AGI demands a leap in creativity and robustness, on par with the best human thinkers, not just competent agents. And he believes that leap is coming—soon.
Wes Roth56,180 Aufrufe • vor 1 Jahr
Greg Brockman says the long-term future is clear: AI will run on its own computer, manage fleets of agents, and solve tasks in parallel. Brockman imagines a workflow where you review your agent’s work in the morning, guiding it like a partner, but models are not yet smart enough for that. He explains that today’s reality is AI helping directly in the terminal or editor, making coding workflows familiar but more efficient.
Wes Roth39,230 Aufrufe • vor 8 Monaten
Eric Schmidt explains that general intelligence—AI with free will and strategic thinking—could emerge in 4–6 years. The next step, superintelligence, will surpass all human intelligence combined. A true test of superintelligence would be an AI proving something humans know is true but cannot comprehend. Schmidt warns this leap could frighten humanity and trigger strong reactions.
Wes Roth44,621 Aufrufe • vor 10 Monaten
Yuval Harari says finance is the perfect domain for AI because it’s purely informational—unlike driving, which deals with the messy physical world. AI will master finance quickly and could even invent financial instruments too complex for humans to understand. Like 20th-century tech shaped both democracies and dictatorships, AI can support very different futures.
Wes Roth42,004 Aufrufe • vor 11 Monaten
Sam Altman said some jobs will completely disappear with AI. He explained young people adapt best, but older workers face more challenges. Altman believes now is the best time in history for young people to create. A single person can now build a billion-dollar company.
Wes Roth34,037 Aufrufe • vor 9 Monaten
Amodei says he feels a responsibility to raise alarms about AI’s labor impact, even while helping to build the technology. He believes the benefits of AI can be huge—but only if we also work to prevent serious harm. With only a few companies leading development, he adds that stopping now wouldn’t halt progress globally—other countries, like China, would quickly take the lead.
Wes Roth35,008 Aufrufe • vor 1 Jahr
Peter Thiel still believes we are in a period of slowed progress—not total stagnation, but reduced speed. From 1750 to 1972, the world saw fast, visible progress: faster ships, trains, cars, planes, and even space travel. Since then, outside of digital tech, most areas have slowed down. Bits (software, internet, AI) have advanced—but the question is: is that enough?
Wes Roth31,712 Aufrufe • vor 11 Monaten
Sundar Pichai acknowledges that right now, AI appears to benefit tech giants and wealthy startups with access to expensive infrastructure. This early concentration of power raises real concerns about whether AI will become another winner-take-all industry. Despite the current landscape, Pichai remains confident that new, dominant AI companies will emerge—just as Google did years after the internet began. He believes the future will be shaped by companies we haven't heard of yet, suggesting there's still room for transformative innovation beyond today’s leaders.
Wes Roth30,757 Aufrufe • vor 1 Jahr