What is the impact of open-source AI?

This resource is a wikidebate, a collaborative effort to gather and organize all arguments on a given issue. It is a tool of argument analysis or pro-and-con analysis. This is not a place to defend your preferred points of view, but original arguments are allowed and welcome. See the Wikidebate guidelines for more.

AI is a powerful tool that can be used to influence people's values, beliefs, and attitudes, but should it? Why or why not?

Open-Source AI is likely to be a positive force in the world

Pro

• Pro Open-Source AI can democratize access to powerful tools which can lead to scientific breakthroughs that can lead to positive impact, like curing diseases.
• Pro Open-Source AI will enable entrepreneurs to develop products and services to make life easier, and save humans precious time doing mundane tasks.
• Pro "Give us back our time!" - People spend a majority of their lives at work. Having an AI tool that can expedite their task and enhances their productivity can give them back time to focus on their personal endeavors.
• Pro It is important that AI used in impactful ways such as in medicine is transparent.
• Pro It is important that AI used in impactful ways such as in medicine is designed for the public interest.
• Pro People's privacy and data can be better be protected if AI is open source and this may even be a requirement.

Con

• Con Outsourcing tasks to AI can result in a narrowing of the scope of paid labor, significantly raising the threshold for hiring talent to perform paid work that otherwise cannot be completed effectively by AI
  • Objection As long as the financial benefits of automation and AI are sufficiently distributed in an authentically ethical way, then automation can help move all of humanity to a post scarcity state of abundance. AI "taking jobs" will be a good thing and not a problem, as long as the increased efficiently and productivity from jobs being "taken" by AI are felt sufficiently by all members of society. For example, an automation tax could be utilized to fund something similar to the Alaskan oil dividends distributed each year (but perhaps something significantly more substantial, depending on to what degree AI and technological automation eliminate jobs. It may be that an open source AI and related technological automation allows all humans to live lives of abundance and post-scarcity; that is, perhaps to live lives of abundance, 99% of all humans only need to "work" 1 to 4 hours per week, and the rest of their time can be spent on chosen pursuits, weather that be volunteering, recreation, spending time with family and friends, traveling, inventing, researching, writing, creating art, hobbies, and/or so on and so forth.
    • Objection There is no evidence or conclusive proof that automation can help move all of humanity to a post scarcity state of abundance, given the natural resource constraints. The four-hour week is arguably an unrealistic fantasy.
      • Objection - "[...] that automation can help move all of humanity to a post scarcity state of abundance, given the natural resource constraints." is a theoretical possibility. There are already efforts to colonize Mars. With an internet search ("solar system sustain a trillion" without quotes) one can see that at some have stated that with space stations that this solar system could sustain a trillion humans. There is no proof that all humans cannot live lives of abundance either; and there is no proof that technological automation combined with ethical capitalism and just laws cannot help all humans to live lives of abundance and reduce the need for human labor to 32, 24, 16, 8, or 4 hours of labor/work per week. This is a theoretical possibility and not really disprovable nor provable until it either happens or never happens after an infinite amount of time.
        • Objection As for "There are already efforts to colonize Mars": no such serious efforts are ongoing; rather, there is an effort to get into a potentially lucrative satellite Internet business.
          • Objection Although Elon Musk has made wrong predictions about feasibility/timing in the past, (like about how fast self driving cars will be functional and common) Elon Musk is the wealthiest human on Earth who has and can allocate billions of dollars in capital for research and development, who has developed several billion dollar companies, and Elon Musk has stated potentially credible plans about how humanity will colonize Mars.
            • Objection The above statements are far from sufficient for the conclusion that Elon Musk genuinely/sincerely intends to colonize Mars (rather than expand rocket making business for various other ends), given the likely conclusion from Is colonization of Mars in this century realistic? and Space Colonization - Pros & Cons, procon.org. One item singled out: If he was serious about the project, he would be running a new version of Biosphere 2 project to ensure preparatory steps other than rocket making. Given Elon Musk's apparent intelligence and ability, it is unlikely he believes the things he is saying about Mars and space colonization.
        • Objection The above are arguably wild science-fiction fantasies with no basis in known reality, especially the idea that the Solar System can sustain a trillion (1,000,000,000,000) humans. See also Is colonization of Mars in this century realistic?. More is probably for a separate debate.
          • Objection You cannot predict distant future. Some stated that planes flying was impossible before that happened. A trillion humans in this solar system is something that in theory could happen in several hundred or several thousand years. To state that this distant theoretical (currently science fiction) possibility is impossible is not really falsifiable nor provable in a any real way. One could have said the same thing about The Space Station 1000 years ago. Regardless, this seems to becoming rather tangential to the original debate topic, "What is the impact of open-source AI?". Perhaps another debate can be started, Can this Solar System potentially support a prosperous human population of over 500 billion humans who can all live lives of abundance and post-scarcity?.
            • Objection One can predict distant future, e.g. that Sun is going to expand and make the Earth uninhabitable.
            • Objection A statement of impossibility is in fact falsifiable/testable: it is refuted once someone makes the thing possible. By contrast, a statement of the form "X will possibly happen in future" cannot be directly falsified/refuted: no amount of X not happening is going to refute the possibility; and that casts doubt on its scientific character (what is not falsifiable, is not scientific).
            • Objection Inference of the form "we did not anticipate planes, therefore anything is possible" is invalid.
            • Objection To analyze the impact of AI in terms of extremely unlikely outcomes is to muddy the waters of the discussion and distract from realistic considerations.
            • Objection A de facto science-fiction article from the website of the grade that slate.com has can hardly be considered a serious argument.
        • Objection The concept of a purely theoretical possibility has almost no cognitive value; it is a theoretical, speculative possibility that we live in a simulation and that the simulation operator decides to halt it and solve many of our problems. There is almost no practical value in these kinds of speculations.
• Con Open-Source AI can democratize access to "catastrophe weapons," meaning it can make it easier to develop chemical and biological weapons, plan kinetic attacks, and deploy cyber attacks.
  • Objection "Regulations would stop it from getting to that point!" - Global institutions would acknowledge the potential hazards posed by open-source AI and take proactive steps to safeguard against the development of such technology reaching critical, catastrophic capabilities. These measures might include rigorous monitoring, recurrent testing, and other security protocols.
• Con Open-Source AI means that hostile foreign nations can access these tools freely, and use them to manipulate and control their populations, such as by generating fake news media to propagandize the public.
  • Objection They can also do so with foreign proprietary AI. If they can't, they can develop it themselves.
  • Objection This depends on which AI soft is made open source.
• Con AI not steerable" -. Current generations of publicly available LLMs are seemingly unable to be censored. Simple prompt hacking routinely get even the most RLHFed models to provide dangerous information.
  • Objection "dangerous information" can also be gained from properly searching the web and the information itself more or less isn't dangerous or isn't the dangerous factor.

External links

• Is keeping AI closed source safer and better for society than open sourcing AI?, largest structured arguments map relating to this subject on Kialo