Of course, there's one aspect of OpenAI's mission I can fully get behind…
ETHICS
-
LLMs and Embeddings: Mimicking Understanding Without True Comprehension
By
–
LLMs mimic understanding through embeddings, representing text and images as numerical arrays for comparison, not comprehension. This concept is central to image synthesis models like Dalle, revealing how machines interpret our image requests.
-
Large Language Models Excel at Numerical Comparisons, Not True Understanding
By
–
Here's the mechanics behind our interactions with machines. Large language models like GPT-4 excel in numerical comparisons, not actual language understanding, as they convert words into math.
-
OpenAI governance challenges and funding: balancing mission with resources
By
–
Of course, if they didn't do these things, maybe OpenAI wouldn't have raised much money at all, and would be a lame duck in trying to achieve their mission. Like I said: governance is hard.
-
OpenAI’s Governance Mistakes: VC Leadership and Profit-Driven Incentives
By
–
I do think OpenAI made many governance mistakes. Having someone from the VC world run it, and having a fintech expert as the CTO, and then hiring people on the basis their compensation would be tied to profits, are all gonna make things even harder!
-
OpenAI Board Addresses AI Safety and Profit-Driven Control Concerns
By
–
I don't know the answers to these questions. But I'm glad the OpenAI board is thinking about them. My view is that previously OpenAI was on a path to reducing safety, because I think having the power of AI controlled by a profit-chasing entity is the worst of all possible worlds
-
Founder Principles: Honesty, Mission Alignment, Board Accountability
By
–
I believe in founders. I believe in being honest and upfront about what you're trying to achieve, and how. I believe that boards should follow through on the mission they've been given.
-
Governance Challenges: Balancing Societal Outcomes Against Profit Incentives
By
–
Also, governance is really, really hard. I've advised many big company boards over many years, and every board I've dealt with has taken their obligations seriously. Trying to maximize societal outcomes over $$$ outcomes is *really* hard to do in our capitalist system.
-
OpenAI Board Refocuses on Mission Despite Disagreement
By
–
My support of the OpenAI board is confusing a lot of people (understandably!) so let me clarify: I don't personally believe in OpenAI's mission. But I still support the board's decision to re-focus on it, because I think that's what boards should do.