Dario Amodei’s AI Predictions Through 2030
EP 82 of The Logan Bartlett Show: Untold stories from tech's inner circle
Very excited to share this episode with one of the most influential people in the AI movement, Dario Amodei (CEO and Co-Founder at Anthropic). We go deep into many topics, but in this letter I’ve broken down his biggest predictions about the future of AI.
Click here to view the episode transcript | Watch on Youtube | Listen on Spotify or Apple Podcasts
✉️ Episode Memo
Dario’s AI Predictions Through 2030:
In 2024, Dario hints we will see crisper, more reliable models that can do longer tasks. These models will likely wow people with their improved capabilities but probably won’t bend our reality too far. Further, in 2024:
Billion dollar models will likely be trained
Models will become more commercially applicable than today
AI’s ability to use tools with improve drastically, making AI-assisted tasks more seamless
Multi-modality will play a big role in advancements (we've seen it from multiple companies in the past few weeks)
Dario predicts that AI’s ability to really invent new science, cure diseases, and make bioweapons will happen no sooner than 2025. He also predicts we’ll see models trained on several billion (maybe even $10B) of investment in GPUs and custom chips in the next 2-3 years.
5-10 years out, we’ll know a lot more about the risks, dangers, and trajectory of potential superintelligent AI models. Here are some of Dario’s bolder predictions:
AI will likely surpass the sum total of human scientific and engineering output
AI models will help us cure cancer, treat diseases, and expend the human lifespan
AI could eventually build Dyson spheres to encircle stars and capture their energy
Dario gives us a 75-90% chance that crazy tech will be built and everything will go fine, but he spends much of his efforts reducing the 10-25% chance that disaster could occur.
Finally, Dario’s advice for developers:
Keep a long-term view and build for 2 years out. Even though AI models may only be able to complete a task 40% of the time today, rapid improvement means they'll probably succeed 80-90% of the time by the time your product is ready to deploy. So take leaps of faith for the expansive possibilities that will appear in 2 years.
⭐ Highlight: Advice for AI Builders
Tiktok failed to load.Enable 3rd party cookies or use another browser
📱 Follow The Show!
If this was helpful to you, please consider forwarding!