Post AThkow0wTYrEy0dAIq by Mastokarl@mastodon.social
(DIR) More posts by Mastokarl@mastodon.social
(DIR) Post #AThhWF19IXFkXY7dAm by ceperez@sigmoid.social
2023-03-17T10:33:54Z
0 likes, 0 repeats
1/n The private biotech industry perhaps has at least 10-20 years more accumulated knowledge than the academic world. A majority do not realize this. This is why nation-states like China and Russia cannot create vaccines like in the West. AI will end up similarly.
(DIR) Post #AThhWLGG5Gu7u9EmYa by ceperez@sigmoid.social
2023-03-17T10:33:54Z
0 likes, 0 repeats
2/n This large disparity in knowledge and know-how is a consequence of the evolutionary nature of biology. This nature also exists in deep learning AI. Evolution's creativity is a consequence of frozen accidents; these accidents cannot be uncovered through first principles.
(DIR) Post #AThhWLp006h3dtUWhs by ceperez@sigmoid.social
2023-03-17T10:33:55Z
0 likes, 0 repeats
3/n Genentech, an early pioneer in genetic engineering, has specially engineered organisms (i.e., mice) that the rest of the world cannot access. Many biotech companies have secret sauce that can only be discovered through experimentation.
(DIR) Post #AThhWaM7xjVmih6BHs by ceperez@sigmoid.social
2023-03-17T10:33:56Z
0 likes, 0 repeats
4/n The same will happen with deep learning AI. We have reached that state in development where firms like OpenAI are keeping their cards very close to their chest. There are scant details on the methods used in ChatGPT or GPT-4. We can only speculate as an outside observer.
(DIR) Post #AThhWb31OFp4rjAR96 by ceperez@sigmoid.social
2023-03-17T10:33:57Z
0 likes, 0 repeats
5/n Most biological material is available to everyone, so it's possible for biotech firms to create similar function genetic material via convergent evolution. In AI, access to the largest models are monopolized by few players.
(DIR) Post #AThhtWKX61ACtXHzE0 by ceperez@sigmoid.social
2023-03-17T10:38:06Z
0 likes, 0 repeats
6/n But surprisingly, open-source models such as #stablediffusion have created their own cottage industry of innovation that has even surpassed closed-source systems. This is because it's unlikely for a private company to be clairvoyant of the entire long tail of use cases.
(DIR) Post #AThhtbOGHa7KaY4XEO by ceperez@sigmoid.social
2023-03-17T10:38:08Z
0 likes, 0 repeats
7/n Furthermore, there is a new debate where private companies like OpenAI argue that open-source AI is unwise because it's too potent. https://www.theverge.com/2023/3/15/23640180/openai-gpt-4-launch-closed-research-ilya-sutskever-interview
(DIR) Post #AThhtj4LjUwWP6okIy by ceperez@sigmoid.social
2023-03-17T10:38:09Z
0 likes, 0 repeats
8/n The most potent biotechnology and the most potent cognitive technology will be in the hands of private corporations. But how is this any different from the present reality that the most potent decentralized coordination is in the hands of private companies (see: money)?
(DIR) Post #AThhvtSQ53ojCZ6rQW by aha@artoot.xyz
2023-03-17T10:38:31Z
0 likes, 0 repeats
@ceperez interesting..somehow an aside, apparently uk government fancies developing a #britgpt - a british oriented ai.hard to tell whether that will go towards a commercial and closed approach - or offer a transparent way of doing #ai
(DIR) Post #AThiFVJbYBIA7MKzVg by ceperez@sigmoid.social
2023-03-17T10:42:05Z
0 likes, 0 repeats
9/n There you have it folks, the most powerful of technologies will always be in the hands of private companies. It has been like this for a very long time (see: central banks). Do not expect that to change in biotechnology or in AI.
(DIR) Post #AThj9dSSSk83aZFCSG by ErikJonker@mastodon.social
2023-03-17T10:52:12Z
0 likes, 0 repeats
@ceperez that's not so clear-cut, nuclear technology is an example, also with chip technologies governments are busy defining boundaries what is allowed or not (ASML in my country). If technology becomes to powerful, dangerous or has a big impact on sovereignty, government will (try to) take control in some form or regulate it.
(DIR) Post #AThkAVZJNfjbMcQiJc by ceperez@sigmoid.social
2023-03-17T11:03:35Z
0 likes, 0 repeats
10/n There is often this argument that everyone in AI knows what others AI groups know. This *was* true before ChatGPT. Knowledge blindness is more likely for technologies of an evolutionary nature. There is a wall of ignorance as a consequence of the Halting problem.
(DIR) Post #AThkAYN0xqgG3LhR44 by ceperez@sigmoid.social
2023-03-17T11:03:36Z
0 likes, 0 repeats
11/n That wall of ignorance becomes a defensive moat when companies finally discover its potency. You see, prior to 2022, deep learning AI was not potent enough. Then a phase transition happened, and all of a sudden we are in an entirely new regime.
(DIR) Post #AThkow0wTYrEy0dAIq by Mastokarl@mastodon.social
2023-03-17T11:10:53Z
0 likes, 0 repeats
@ceperez Thanks, excellent points. In AI, it’s also difficult for open source etc. to keep up because training these models costs an enormous amount of money for hardware and the energy to run it.Had to laugh about the comment from the openAI guy that a certain technology is so powerful that it is too dangerous to opensource it.
(DIR) Post #AThscWYG7QHu9co7mK by ceperez@sigmoid.social
2023-03-17T12:38:17Z
0 likes, 0 repeats
12/n There are "hyperobjects" in this world that contain know-how and knowledge that are *do not* have mobility. They are affixed in the development and evolution of an "organism" (i.e., a living thing, an AI, a language, a corporation, a community, a city, a nation-state,… https://twitter.com/i/web/status/1636708143840346113