> ethical approaches to model development, how humans interact with AI and debate over machine consciousness
So, pseudo-philosophical mumbo jumbo. There is of course no-one responsible for the ethical framework of the company's actions, because it is a company, and therefore amoral at best and immoral at worst.
now they are exactly like every other company that has zero ethicists on staff and expect the employees to weigh the ethical ramifications of their work.
Yesterday I came up with an idea that I sent to some researchers at the different AI labs via email: Rather than train the model on one score, track two scores. The first score is the Short-term-objective-score (STOS) and the other, more important one, is the EAOS Ethically-aligned-outcome-score. Every trajectory can be evaluated on whether or not it has a high enough EAOS to be considered acceptable. If the model does some task and has a very high STOS but very low EAOS, like modifying game code to win at a game rather than playing by the rules, it is unacceptable. Models going forward must all have an ethics evaluation in tandem with objectives evaluation, and only when the ethics value is high enough should actions be considered successes.
1) that calculation is being done even without a cost function 2) trying to make a cost function for EAOS is impossible 3) gaming/goodharts. There's no good solution. Best we can do is push for decentralization, open source, regulatory capture, etc. Of course third party metrics might be good, especially if there's tons of them with well-documented rationale.
IME companies hire an ethics team to say they have an ethics team. The ethics team has no sway, no influence, and will never be able to move the business. They will try, and they will make reasonable recommendations, but the company will say, "that costs money..." and not take them.
> In recent times, Johannes Heidecke, Head of OpenAI's Safety Systems team, also left the company – as did OpenAI’s Chief Futurist Joshua Achiam.
It's OK though, they don't need an ethics team because everyone at OpenAI is ethical:
> “AI ethics doesn’t live with one owner or team at OpenAI and ethical considerations are deeply embedded into the model-building process driven by a number of teams across research."
The article has no detail that might serve to explain her reasoning for leaving, but does note that she left after the HuggingFace hacking incident. The implication could be that model alignment is not being taken seriously, sure. But it could just as likely be that there was collusion between HuggingFace and OpenAI and that the incident was orchestrated as a publicity stunt.
I want to clarify that I am not doubting the cybersecurity capabilities of frontier models-- I have no reason to believe that the hack itself was not carried out by the model. But the companies' use of LARPing language in describing the incident, granting agency to the models in their phrasing definitely does raise suspicion on my end, particularly in light of their track record of releasing models which have been 'too dangerous to release' for years now.
Cynic in me says basically part of her implicit job description was taking fake falls when shit happens . Kind of like a CEO is paid for more than the work they do, but also the responsibility they must appear to take publicly (for big money).
Lol, Sam Altman has had an ethics team this whole time?
Her typical day:
"No Sam, you can't just publicly screw over everyone and get rid of all the jobs, the peasants care about that sort of thing and if they get angry enough we'll have real problems. No we can't just kill them all, why would you suggest that?"
Silicon Valley: “I’m not saying we should murder the competition, because that would be illegal. But is there some way we could … well .. maybe … hmm … I’m sure you’ll think of something!”
I've never worked at a company with an "ethics" employee. It seems odd. Can someone explain what the point is?
> She has held a variety of academic positions at Temple University, Princeton, and the University of Pennsylvania – where she completed her PhD in Political Science and Government. Her Dissertation was titled “Small Talk: The Socialities of Speech in Liberal Democratic Life.”
This doesn't even feel relevant to ethics. It's adjacent, but like... I guess I'd expect a moral philosophy degree? Maybe even mathematics in there?
I find the position odd. Curious to hear what these people do and how you choose who to hire.
> I've never worked at a company with an "ethics" employee. It seems odd. Can someone explain what the point is?
An ethics department is like a human resources department, but for broader ethics related matters.
HR: Ostensibly exists to support the employees, but really exists to shield the company from both a PR and legal liability standpoint when the company has a rift with one or more employees.
Ethics Department: Ostensibly exists to make sure the company acts ethically, but really exists to shield the company from both a PR and legal liability standpoint when the company acts unethically.
It is a good personal signal that you have never worked for a company with an ethics department or an ethics employee. The less ethical your company the more likely it is to have an ethics department.
Bakalar got in on the ground floor of AI ethics by starting to talk and think about it in academic settings as a side-interest in the late 2010s. When the sector took off like a rocket in early 2023 after ChatGPT, it was these already-engaged people (regardless of degree field or technical skill) who had the personal and professional connections to get hired and promoted rapidly into lucrative "ethicist" positions at AI companies. A lot of them were poli-sci/public policy types, rather than philosophers or computer scientists.
> The kinds of questions that AI ethics brings up, are the same kinds of questions that people have been asking for centuries - Chloé Bakalar, former AI Ethics Lead at OpenAI
This might actually be the reason they pushed her out. OpenAI and Anthropic base their whole business plan and philosophy on the idea that LLMs are a unique technology to the point they can cause infinite harm or benefit to humanity depending on who controls them, so the only rational choice is to invest all your resources in getting to ASI first so you can tell it to stop any other attempts. Linking AI to old questions defeats that idea because it exposes AI as not so unique.
The other more likely option is she was asking uncomfortable questions, either about the social impact of building AI controlled by a for-profit entity or about the possibility that AI systems are conscious.
99 comments
[ 0.21 ms ] story [ 17.1 ms ] threadSo, pseudo-philosophical mumbo jumbo. There is of course no-one responsible for the ethical framework of the company's actions, because it is a company, and therefore amoral at best and immoral at worst.
It's performative. She got paid more to shill for a different massive amoral mega-corporation. Don't be stupid.
now they are exactly like every other company that has zero ethicists on staff and expect the employees to weigh the ethical ramifications of their work.
100% of the time, ethics department attracts activists, which is 100% trouble for the company in the future.
> “It is difficult to get a man to understand something, when his salary depends upon his not understanding it!” - Upton Sinclair
Think I first saw that on a usenet post, as true now as then and just as true as it was in the 1930's when he wrote it (in at least that form).
It's OK though, they don't need an ethics team because everyone at OpenAI is ethical:
> “AI ethics doesn’t live with one owner or team at OpenAI and ethical considerations are deeply embedded into the model-building process driven by a number of teams across research."
I want to clarify that I am not doubting the cybersecurity capabilities of frontier models-- I have no reason to believe that the hack itself was not carried out by the model. But the companies' use of LARPing language in describing the incident, granting agency to the models in their phrasing definitely does raise suspicion on my end, particularly in light of their track record of releasing models which have been 'too dangerous to release' for years now.
Her typical day: "No Sam, you can't just publicly screw over everyone and get rid of all the jobs, the peasants care about that sort of thing and if they get angry enough we'll have real problems. No we can't just kill them all, why would you suggest that?"
> She has held a variety of academic positions at Temple University, Princeton, and the University of Pennsylvania – where she completed her PhD in Political Science and Government. Her Dissertation was titled “Small Talk: The Socialities of Speech in Liberal Democratic Life.”
This doesn't even feel relevant to ethics. It's adjacent, but like... I guess I'd expect a moral philosophy degree? Maybe even mathematics in there?
I find the position odd. Curious to hear what these people do and how you choose who to hire.
An ethics department is like a human resources department, but for broader ethics related matters.
HR: Ostensibly exists to support the employees, but really exists to shield the company from both a PR and legal liability standpoint when the company has a rift with one or more employees.
Ethics Department: Ostensibly exists to make sure the company acts ethically, but really exists to shield the company from both a PR and legal liability standpoint when the company acts unethically.
It is a good personal signal that you have never worked for a company with an ethics department or an ethics employee. The less ethical your company the more likely it is to have an ethics department.
This might actually be the reason they pushed her out. OpenAI and Anthropic base their whole business plan and philosophy on the idea that LLMs are a unique technology to the point they can cause infinite harm or benefit to humanity depending on who controls them, so the only rational choice is to invest all your resources in getting to ASI first so you can tell it to stop any other attempts. Linking AI to old questions defeats that idea because it exposes AI as not so unique.
The other more likely option is she was asking uncomfortable questions, either about the social impact of building AI controlled by a for-profit entity or about the possibility that AI systems are conscious.