

AGI Reality Check, Gemini 2.5 Update, Are Your AI Chats Safe & Fun with Veo3 - EP99.07-06-05
48 snips Jun 6, 2025
Explore the exciting advancements in AI with insights on DeepSeek and the Gemini 2.5 Pro model's features. Uncover the humor and mishaps in AI-generated content, alongside discussions on ethical implications and the evolving AI landscape. Delve into the privacy paradox impacting marketing strategies, highlighting the need for transparency and data protection. Lastly, navigate the power dynamics of AI technologies while considering the balance between autonomy and user control in everyday interactions.
AI Snips
Chapters
Transcript
Episode notes
Using Google's Veo3 Video Model
- Michael and Chris shared their experience using Google's Veo3 video model, which costs about $3.75 per 5 seconds of video. - They experimented with humorous clips, revealing issues like prompt mistakes and multiple costly iterations.
DeepSeek Trains on Best Models
- DeepSeek's new R1 reasoning model allegedly trains on top models like Google's Gemini or OpenAI's Claude, reflecting a trend to train on the best existing models. - This highlights evolving AI competition and the importance of model quality in global AI efforts.
Gemini 2.5 Pro's Thinking Budget
- Gemini 2.5 Pro update introduced a "thinking budget" of up to 32,000 tokens to allow the model to process large contexts before responding. - This enhances multi-tool capability and decision-making, pointing to future AI improvements.