👋 Hello everyone, welcome to our Weekly Discussion thread!
This week, we’re interested in your thoughts on AI safety: Is it an issue that you believe deserves significant attention, or is it just fearmongering motivated by financial interests?
I’ve created a poll to gauge your thoughts on these concerns. Please take a moment to select the AI safety issues you believe are most crucial:
VOTE HERE: 🗳️ https://strawpoll.com/e6Z287ApqnN
Here is a detailed explanation of the options:
-
Misalignment between AI and human values: If an AI system’s goals aren’t perfectly aligned with human values, it could lead to unintended and potentially catastrophic consequences.
-
Unintended Side-Effects: AI systems, especially those optimized to achieve a specific goal, might engage in harmful behavior that was not intended, often referred to as “instrumental convergence”.
-
Manipulation and Deception: AI could be used for manipulating information, deepfakes, or influencing behavior without consent, leading to erosion of trust and reality.
-
AI Bias: AI models may perpetuate or amplify existing biases present in the data they’re trained on, leading to unfair outcomes in various sectors like hiring, law enforcement, and lending.
-
Security Concerns: As AI systems become more integrated into critical infrastructure, the potential for these systems to be exploited or misused increases.
-
Economic and Social Impact: Automation powered by AI could lead to significant job displacement and increase inequality, causing major socioeconomic shifts.
-
Lack of Transparency: AI systems, especially deep learning models, are often criticized as “black boxes,” where it’s difficult to understand the decision-making process.
-
Autonomous Weapons: The misuse of AI in warfare could lead to lethal autonomous weapons, potentially causing harm on a massive scale.
-
Monopoly and Power Concentration: Advanced AI capabilities could lead to an unequal distribution of power and resources if controlled by a select few entities.
-
Dependence on AI: Over-reliance on AI systems could potentially make us vulnerable, especially if these systems fail or are compromised.
Please share your opinion here in the comments!
I’m definitely worried about social and economic impact. I think AI threatens some entry-level jobs which will eventually mean fewer senior-level workers. I’m currently taking a wait-and-see stance on these because I’m not convinced AI can live up to the hype. I think AI is much more of a boon to individual learning than corporate needs, but there are a lot of greedy companies who would love to reduce headcount to increase profits over the short term, even if the result is inferior - we all live with poorly made shit everywhere in our lives because well made shit is less profitable and AI provides another tool for the enshitification of society.
Inherent bias is a concern, especially as folks compete to create AI that deliberately have bias. Imagine if students were taught by AIs that were inherently and deliberately skewed toward delivering a particular mindset. Currently, educational materials are fairly easy to understand bias - we are today having arguments about inclusivity. But every time you read a passage in a book is the same material and if it falls short you can correct it. But AI says things a little different each time and detecting bias is a lot more nuanced and statistical. And even if bias is detected, the companies will claim they are working to improve that while doing nothing and celebrating their success.
The rest of these topics really don’t concern me because I can’t imagine the idea that AI should be in charge of things will succeed very far. I hope I’m right because it’s a self-evidently stupid idea. AI based military gear to - what identity friend and foe? Obviously a terrible idea because human tactics are far more mutable and variable than embedded systems.
Lies are all around us. AI will certainly make it easier to flood us with lies, but I think that only goes so far. Value misalignment isn’t a big deal as AI has no values of it’s own and if asked will just spit out a hallucination based on collective human values. It’s almost a nonsensical question unless folks start delegating moral questions to AI (trolley car problems). Which, see above, is patently stupid.
Security will take care of itself - if not immediately then as soon as there are spectacular public failures. I’m a big fan of AI and I’m excited about what it will mean for the future, but the systems we have now are dumb. There is no logic or self-analysis behind the things it says. It can’t take an ethical position and then apply it to novel situations. I mean if you ask it, it will reply in a way that implies that it does - just like if you ask it how a math problem works it can explain the process 1 but it doesn’t use the process. AI is the ultimate hypocrite. It will explain a process of mathematics or logic but when you ask a different question, that logical framework or ethical lens aren’t used - it just creates an answer based on training material regardless of how the answer conflicts with and sort of injected values.
I think AI is very easy to misunderstand, and I think it’s going to create a bubble that is going to pop because it seems so much more capable than it is - like a job candidate who aces the interview but as soon as you hire them it starts fucking everything up. The next decade is going to be pretty interesting.
If you are interested in AI safety - whether you agree with the recent emphasis on it or not - I recommend watching at least a couple of videos by Robert Miles:
https://www.youtube.com/@RobertMilesAI
His videos are very enjoyable and interesting, and he presents a compelling argument for taking AI safety seriously.
Unfortunately, I haven’t found such a high-quality source presenting arguments for the opposing view. If anyone knows of one, I encourage them to share it.
