The speaker discusses the recent advancements in AI, specifically highlighting two new technologies: Bard and Gen Avatar 2.0.
Bard, he explains, is an AI that can analyze images and provide information about them. For example, Bard can identify characters in an image and suggest relevant YouTube videos. The speaker demonstrates this by showing an image of a castle and asking Bard to provide information about it. However, the speaker also points out some limitations of Bard, such as its inability to accurately answer mathematical problems and its tendency to generate hallucinations in its responses.
The speaker then moves on to Gen Avatar 2.0, a technology that can create realistic images of people in different languages. He demonstrates this by showing images of a character, Oppenheimer, in different languages. However, he notes that while this technology is promising, it still has some limitations. For example, it can only be used for translation and not for generating speech without a consent form.
The speaker concludes by discussing the Open AI Red Teaming Network, a platform where domain experts can contribute to the development of future AI technologies. He encourages viewers who have expertise in various fields to join the network.
1. The speaker is discussing a new AI image technique that has revolutionized translation and dubbing.
2. The speaker has tested the new AI, now referred to as "Bard", in various ways.
3. The speaker plans to cover how anyone with expertise in diverse fields could contribute to the development of the technology.
4. The speaker has used Bard to analyze images and provide information about historical figures.
5. Bard can recommend YouTube videos based on the content of an image.
6. The speaker has used Bard to analyze a travel image and provide information about travel to the location in the image.
7. Bard can search Google Drive for documents and summarize them in a Shakespearean sonnet.
8. The speaker has used Bard to read Gmail messages and generate feedback.
9. Bard can sometimes generate false feedback.
10. Bard can recognize images and provide relevant information.
11. Bard can sometimes provide inaccurate information, such as figures for gross revenue.
12. Bard can provide information in multiple languages.
13. The speaker predicts that the technology will be able to rewrite historical footage in different languages.
14. The speaker notes that the technology only works for translation and requires consent for speech synthesis.
15. The speaker predicts that the technology will be able to manipulate political elections with deep fake imagery, possibly as early as 2025.
16. The speaker has used the AI to generate images and turn them into videos.
17. The speaker has used the AI to create bold text in Adobe Express and then upload the output to Runway gen 2 for generation.
18. The speaker has used the AI to generate images with different prompts.
19. The speaker has used the AI to generate images with different seed values for different outputs.
20. The speaker has used the AI to generate 3D images.
21. The speaker has announced the Open AI Red Teaming Network, where domain experts can join and get paid.