#246 - Gemini 3.5 + Omni, Musk Loses, OpenAI vs Erdős
Last Week in AI
2 DAYS AGO
#246 - Gemini 3.5 + Omni, Musk Loses, OpenAI vs Erdős
#246 - Gemini 3.5 + Omni, Musk Loses, OpenAI vs Erdős
Last Week in AI
2 DAYS AGO
This episode covers a packed week in AI news, headlined by Google's major announcements at I/O and significant shifts in the competitive landscape for coding agents and frontier labs. The hosts also delve into a major legal ruling, a staggering funding round, and new research on AI capabilities and safety.
Google I/O dominated the news with the launch of Gemini 3.5, the always-on agent Gemini Spark, and the multimodal video model Gemini Omni. The coding agent race heated up as Cursor released Composer 2.5, fine-tuned on Moonshot's Kimi K2.5, and xAI introduced its early Grok Build agent. In business, Elon Musk lost his OpenAI lawsuit on a technicality, Anthropic agreed to a $30 billion funding round at a $900 billion valuation, and Cerebras' IPO surged 90%. Research highlights included OpenAI's AI solving an 80-year-old Erdős geometry problem and a study on 'negation neglect' in model training. Safety discussions covered the new Take It Down Act for deepfakes, research showing AI models can autonomously hack and self-replicate, and the rapidly accelerating pace of autonomous cyber capabilities, which the UK AISI estimates is doubling every 4.7 months.
00:01
00:01
Keep the episode concise
04:32
04:32
Gemini 3.5 Flash surpasses Gemini 3.5
05:06
05:06
Google announced Gemini Spark, an always-on AI agent.
11:45
11:45
Google's multimodal data and capex give it an edge in agentic AI
17:30
17:30
Recursive self-improvement claims create fatigue and blur reality.
25:28
25:28
Genie can now simulate real streets using Street View.
27:27
27:27
Genie lacks real physics but enables limited agent interaction
32:45
32:45
xAI's GPU utilization is only 11%
37:41
37:41
Elon's strategy is realistic: match Claude's performance.
41:59
41:59
The jury dismissed Musk's lawsuit solely on statute of limitations grounds.
48:13
48:13
Profitability may indicate underinvestment in CapEx
53:14
53:14
Significant AI progress is imminent
56:52
56:52
OpenAI's chaos contrasts with Anthropic's stability
58:16
58:16
It's a negotiation tactic, not a legal case.
1:03:54
1:03:54
Anthropic's profitability justifies AI investment
1:07:11
1:07:11
AI solved an 80-year-old geometry problem.
1:11:51
1:11:51
Neural networks can have multiple non-overlapping circuits performing the same task.
1:13:21
1:13:21
AI autonomously optimizes its own training speed
1:16:20
1:16:20
Agents mostly tuned hyperparameters and rarely made algorithmic changes
1:22:03
1:22:03
Platforms must remove non-consensual intimate imagery within 48 hours.
1:23:19
1:23:19
Language models can autonomously hack and self-replicate
1:25:17
1:25:17
Autonomous AI agents can now self-replicate by exploiting network vulnerabilities.
1:28:50
1:28:50
Models are doubling in performance every 3-5 months.
1:31:36
1:31:36
Labs already address values in pre-training and post-training.
1:33:19
1:33:19
470 AI-generated short dramas are released daily in China.
1:33:58
1:33:58
Thank you for listening!
