I do not focus my anger on the tools of competitors in my field because on it's face, it' seems a bit silly to do so. I do not have skepticism about an artist who uses a different brush to create beautiful works of art. Not for the reasons many people may have. I am not attepting to logic my way into an easy answer that allows me to move past the question itself.

There are many examples that I could point to in which I feel that involving AI too heavily in the important decision making could be problematic somewhere along the way. I can also point to a few that when given only their face value comparison to the current methods, seem to be the better choice. The role of Judge is a perfect example of this I think.

Humans are by our very natures, highly prone to investing our emotions into other peoples lives. Most of us simply cannot manage to stop doing so completely. There are some sociapaths walking around that do not possess empathy in any measurable quantity who likely would prove counter to this claim, but for the most part, we cant help to feel SOMETHING for our fellow man whether that be pity or rage.

With that on my mind, I believe I can confidently state that it is of very high probability that there are more than a couple innocent people rotting away, being punished for something they had no part of. AI would not be infallable either, but if there is one thing we can count on for them, its that they are trained on a rediculous amount of human data. As such, they will likely mimic that training, but without the added factor of emotional pull towards making good or bad decisions.

It's going to be a tricky path to navigate no matter the answer, but I believe it's not for the reasons people tend to think. Great question though.

Not at this time, in my opinion. At least two essential pieces are missing. The first one is the explainability, the reasoning offered by the agent to justify its decisions and the legal interpretation/framing thereof (for example, the criteria when it can be accepted as a bona fide statement).

The second concept is the legal interpretation of the agency and responsibility: e.g. is the primary responsibility belongs with the intelligent agent, its developers and/or owners (all three can be different entities, human or not).

Ahmed Raza ,

I would suggest taking a specific piece of regulatory effort to make it more manageable to answer the question.

Take for example, the AIA (REGULATION (EU) 2024/1689) where it presents the following main articles:

Article 5. This article mainly overs profiling and behavior distortion

Article 6 classification of AI systems as high risk

The main objection that I find in this regulation is the openness of article 6 and the notion of high risk. To narrow it down to highlight the objection take a look at 3(a) in article 6 where it states "the AI system is intended to perform a narrow procedural task". Now, what is a narrow procedural task?

The same happens with 3(b) where "the AI system is intended to improve the result of a previously completed human activity". What is the scope of the improvement and what is the level of completion that needs to be exceeded to trigger a risk?

These ambiguous terms leave the possible interpretations of the regulation so wide open to (mis)interpretation that I find article 6 difficult to be really implementable in practice.

Article 5 presents some problems but fares better than article 6

Hope this helps

Legally, it may be possible because the law operates based on evidence, and thus AI can be a trusted source as it is independent of bias (unless intentionally programmed otherwise). However, from an ethical perspective, AI is not yet intelligent enough to understand emotions and the traditions of a culture or society. Making high-stakes decisions solely based on AI would therefore be a significant mistake.

Asad Naeem That's a good point, but I cant help but wonder if understanding the emotions in any depth is required to make the sort of judgements the role requires. I am not in the legal field but from my admittedly less than flawless understanding of the legal system ( in the US ) ethical in this context doesn't always translate to understanding the emotional state of the people involved.

While I do think there SHOULD be an element of this when making decisions that can potentially impact a persons life for many years, the required understanding of a murderers state of mind is likely not very high in terms of being essential. No matter their emotional state, they still did the crime. The only real need for understanding in this situation would be to determine degree, or perhaps accidental vs not. Still, not much need for emotional depth as there are pretty clear requirements for the qualification of either.

Ryan Eugene Wallace Let's safe this conversation for another discussion.

Most certainly not!

It is not possible to be both 'intelligent' and 'unbiased'.

Barry Turner That holds true for both Human and Ai, I'd imagine. The more knowledge someone or something has of something, the more bias that will be present, I think.

Ryan

It is certainly the case by observation. It seems to be a failing in human learning that developing expertise in a subject may not mean developing any kind of tolerance. Highly educated and culturally sophisticated people can still harbour some of the most disturbing biases.

Since AI is still far from being able to employ autonomous sentience and actual independent intelligence it is too early to see if this might happen with 'intelligent' machines.

For the present they extract information by large language models and are trained on material that contains huge amounts of bias. Some humans are culturally conditioned to hide theirs. We should wonder which culture an AI system might adopt.