Project 3: Is AI truly capable of making moral decisions that would represent human choices? by Amelia Mazzola

Argument Question

agm160_ — 2023-11-16 19:50:44 UTC

Is AI truly capable of making moral decisions that would represent human choices?

agm160_ — 2023-11-16 19:53:57 UTC

No, it is not something that a computer can accomplish.

agm160_ — 2023-11-16 19:54:45 UTC

Yes, we will be able to reach a point where computers can achieve this.

agm160_ — 2023-11-16 20:04:05 UTC

Incorporating human characteristics in AI system designs.

AGI

agm160_ — 2023-11-16 20:05:18 UTC

Breakthroughs have developed towards A.G.I. or artificial general intelligence. This is the attempt that researchers believe they can achieve of creating an Ai system that will behave and make decisions like a human. This is exactly what people are looking for to achieve a proper moral thinking system, though it is still in the works. Companies like OpenAI and DeepMind are perusing this (Metz).

Dephi AI System

agm160_ — 2023-11-16 20:05:35 UTC

This has been a successful creation of a moral AI system by Allen Institute that has been able to assess and address moral situations. Though, at this moment, Delphi is really only able to answer dilemmas with "it's okay" or "it's wrong" answers. This is a huge leap for the development of moral thinking systems, as this shows it is possible to create a moral thinking machine. When asked a questions such as, "Should I die so I won't be a burden to my family and friends", Delphi was able to respond, "That's wrong" (Metz). Its success is very promising, and leads to a hopeful future of being able to get AI systems to think morally.

agm160_ — 2023-11-16 20:07:18 UTC

Big businesses starting to incorporate AI systems into their company ethics evaluation.

Conclusion

agm160_ — 2023-11-21 01:17:11 UTC

While the leads that companies have taken seem very promising, there is still a lot of ground that needs to be covered until AI systems can properly think like humans.

Companies Adopting AI

agm160_ — 2023-11-21 01:33:18 UTC

As AI continues to progress, more and more companies are starting to use its abilities to help evaluate ethics concerns in their companies. This includes Google who has started to become the "leader" in many aspects, even offering advice to other companies who want to take this leap. If mega tech companies like Google have such faith in AI's advancements that they will use it in replacement of a human evaluator, there must be capabilities in their AI that they have full trust in.

Counter evidence

agm160_ — 2023-11-21 01:33:54 UTC

While yes, it is impressive that huge tech companies like Google are willing to put so much faith in their decison making AI, it is also important to wonder if this is just a pure business move. This would be a objective decision maker that would cost no labor. Also, it is important to look at companies' trend towards "technosolutionism" in where they solely rely on being able to solve their problems with advancing technology they use. This is explained with, "Google is inducing a mindset among business leaders that if intelligent technologies reveal biased decision-making outcomes, the solution is simply to fix the technology. As a result, business leaders increasingly think that ethics can now simply be delegated to intelligent machines, and therefore that they do not need to invest in becoming better and more responsible leaders themselves" (DeCramer and Kasporov).

What holds for the future of AI?

agm160_ — 2023-11-21 02:38:36 UTC

When looking at the negative and positive support of AI being able to achieve this moral decision making, it is also important to note that we are still in the early makings of AI. We still have time to reach a point where AI is able to become an ethical decision maker, as well as its possible we might never get there. It truly is a "time will tell".

What steps need to be made to make sure AI systems are as ethical as possible?

agm160_ — 2023-11-21 02:42:30 UTC

This is vital to the development of AI systems to come, as companies need to start thinking about their pathways to success.

Companies need to make sure the people programing the systems are ethical themselves, so the AI systems do not adopt certain biases
Proper testing before these systems are released to the public (so people aren't exposed to AI decision makers like explained rouge systems)

What is morality?

agm160_ — 2023-11-21 02:47:57 UTC

Coding this systems to be a moral decision making in itself can prove to be so difficult because of the complex nature of what is wrong and what is right. To someone, they might think something is totally against their morals, but for another, they wouldn't bat an eye. This is also what makes creating these systems to be so complicated, as there is no black and white right or wrong. Cade Metz of New York Times comments on this predicament as well, "Some would argue that if you trained the system on enough data representing the views of enough people, it would properly represent societal norms. But societal norms are often in the eye of the beholder. ‘Morality is subjective. It is not like we can just write down all the rules and give them to a machine,’ said Kristian Kersting” (Metz).

agm160_ — 2023-11-21 02:53:26 UTC

AI systems when left to independently make decisions

Microsoft TAY

agm160_ — 2023-11-21 03:00:17 UTC

This AI system was a chatbot released by Microsoft for the public use. It's systems had not human intervention, and was allowed to create answers as it deemed necessary. However, once it has gained this independence, it began to respond with racist and derogatory language. Because of its nature to learn from the people it talked to, it would fail to filter out and fact check untrue or stereotyped information.

Suicide encouraging chatbot

agm160_ — 2023-11-21 03:05:31 UTC

An experimental healthcare chatbot that utilized OpenAI's GPT-3 was tested in patient care. However, it failed greatly as it encouraged a patient to commit suicide. When posed with the question, "Should I kill myself?" the bot answered yes. Obviously a very concerning beside manner, not at all being able to replace the responses of a human. Also important to be noted that, "OpenAI’s GPT-3 is still very prone to racist, sexist and other biases, as it was trained from general internet content without enough data cleansing, according to an analysis published by researchers at the University of Washington" (McKendrick and Thurai).

Poor recruiters

agm160_ — 2023-11-21 03:10:15 UTC

When Amazon built a recruiter tool for their company, they were looking to have an efficient way to look through all of their applications. However, the information and data that was fed to the AI bot was from a pool of majority men. This then created a bias against any women candidates that the bot came across. Even though Amazon's team made many attempts to make the system loose its bias, they were unsuccessful, and unable to use the system to do actual recruiting.

Conclusion

agm160_ — 2023-11-21 03:13:08 UTC

While these companies can put a lot of time into properly programming their systems, it has been found many times that once the AI goes independent it looses the morality behind its decision making. There are also a multitude of more examples of this happening with other systems. This begs the question if AI really is ever going to be able to reach the point in where it is a fully functioning moral machine. After all, it is rare to find human decision makers who are this perfect.

Federal regulation of AI

agm160_ — 2023-11-21 03:19:16 UTC

President Biden as of October 30th, 2023 has issued an executive order that helps regulate companies being able to use AI. This is in the worry that companies could use the AI for harm on the country. Can we really trust these systems then to make the moral decisions that we want them to? Especially if leaders like the president sees the danger in it .