Extremely interesting. Stable Diffusion XL says nothing to me, so in a sense this is bad journalism. If I understood it correctly then generating a small picture for my PPT is equal to driving 1-5km with my car (?). Thats fairly bad, and people would probably use these less if we knew.

Cognitive architecture captures the way a bigger service or process is split up in different interrelated tasks to be carried out by LLMs and traditional software, relating often to org design. It is a term I first encountered in a LangChain blog post here: blog.langchain.dev/openais-bet-... (2/2)

Most of my research and teaching is now related to generative AI. Looking at all the elements that make up the "stack", the most likely source of competitive advantage are the green boxes -- the domain of management and organizing. Cognitive architecture is also interesting... (1/2)

I really have to agree, 100%! 😉

So does this mean that Israel denying the supplies to Gaza is simply causing civilian suffering without accomplishing anything?

For real benefit, batch processing w GPT3.5 at this point. Extracting sentences related to topic(s) of interest from a large archival dataset.
I also used GPT4 as a "research companion" to create second order categories from quotes. It does this surprisingly well, but legitimacy might be an issue.

I am working on this, looking at using GPT with archival data (news articles).

I am both modes at once, "I really loved it" + total rewrite. 😜

Cool! I am working on how to integrate large language models to research & teaching, but so far ignored service. There is one additional super important element: business schools can study & teach ways for companies to benefit from LLMs in an effective and responsible manner.

It seems that every week I have a conversation with someone about the impact of generative AI on professional work and how business schools should start to figure out what skills our graduates will actually need.
Not sure how this "figuring out" will ever start though. #llms #bschools