ChatGPT Search

I think you're conflating separate phenomena:

* Boeing is a consequence of the "Jack Welch" effect - gutting the core in service of short term gains for stock-holders.

* The MBA type, typified by John Sculley at Apple is about calcifying the current offerings presuming market segments supported by historicals with predictable demand. This works well for defensives such as utilities, consumer products and health care but not for markets with dynamic consumer relationships such as technology.

* The Google Cloud example is the Xerox Parc phenomena. Xerox was organizationally structured for investment payoffs only characteristically similar to their mainline products thus they couldn't properly allocate resources to things, such as desktop computing, with different kinds of curves. This is similar to how the franchise retailer Blockbuster so slowly responded to the centralized mail-order subscription Netflix. The institutional structure is only-so-flexible. This is similar to Conway's Law.

* The "ruin everything else" is a generalized form of a "brand extension failure". Examples include Harley Davidson perfume, Bic underwear, McDonalds Pizza, and Heinz cleaning vinegar - an over-leveraged commitment to a wildly successful core offering makes other ventures impossible.

This is not that. It's yet something else. Abstractly it's "X is a wild success, let's make Y another X instead of working on X+1"

Organizations suffer from varying degrees of ailments and they can create codependencies making the unraveling hard. Often it devolves into politics of power brokers with the company's survival dependent on the competency of the influential instead of the influence of the competent. A brutal struggle to control a sinking ship.

The crisis of the third century happens every day.

> I’m not sure Facebook fits in considering they at least managed to get some other products along the way

can you name these products?

As with several others in this thread, I've been using Kagi's GPT-aided search, FastGPT, for much of this year, figuring I should probably dip my toes into what LLMs are making possible, as well as their limits.

For someone who's used online keyword-based search since the 1980s (computerised library catalogues, at the time), it's jarring for me to get over the distinction of querying for documents (old school) as opposed to asking direct questions, but that's precisely what LLM-based GPTs facilitate.

And as I'd noted this past June, it's a sea change in online search:

[O]ne of the upsides of GPTs / LLMs is that they provide direct answers to questions, though those answers may be hallucinations or generalisations. Even then, the directness is refreshing, though I expect it to also get polluted rapidly through both AI SEO manipulation and advertising / general enshittification of AI engines.

https://toot.cat/@dredmorbius/112577405443953191>

I generally rely on Kagi's LLM for various reasons, but foremost is that it relies on current Web search and cites its findings specifically, which makes validating responses and detecting hallucinations far easier. ChatGPT specifically would hallucinate not only its responses but the citations it provided when those were requested, which curbed my enthusiasm greatly. It'll be interesting to see how its search-oriented offering fares.

I strongly agree with Temporal's excellent observation that this is, at least for the moment, a strong shift of the Web back to serving readers rather than advertisers (first and foremost) and publishers (distant second): https://news.ycombinator.com/item?id=42011414>.

What I strongly suspect is that any successful GPT search tools will be rapidly engrossed by existing search monopolies. For those who defend the "free market" on the basis that competitors can emerge, the countervailing force of mergers and acquisitions must be noted, as well as the fact that these almost always effectively destroy that competitive potential, at least over the past half century or so.

And among those other uses, I've found that generic LLMs are surprisingly effective as translation engines though they were not specifically intended for this role. That includes "translating" to various dialects (which may be easier to proof or check).

The results seem to be better for strongly represented languages (e.g., English, Spanish, German, French), less so for those which may be less prevalent online (e.g., Yoruba):

https://toot.cat/@dredmorbius/113283430170325203>

Traditional search can become "spamming text" nowadays because search engines like Google are quite broken and are trying to do too many things at once. I like to think that LLM-based search may be better for direct questions but traditional search is better for search queries, akin to a version of grep for the web. If that is what you need, then traditional search is better. But these are different use cases, in my view, and it is easy to confuse the two when the only interface is a single search box that accepts both kinds of queries.

One issue is that Google and other search engines do not really have much of a query language anymore and they have largely moved away from the idea that you are searching for strings in a page (like the mental model of using grep). I kinda wish that modern search wasn't so overloaded and just stuck to a clearer approach akin to grep. Other specialty search engines have much more concrete query languages and it is much clearer what you are doing when you search a query. Consider JSTOR [1] or ProQuest [2], for example. Both have proximity operators, which are extremely useful when searching large numbers of documents for narrow concepts. I wish Google or other search engines like Kagi would have proximity operators or just more operators in general. That makes it much clearer what you are in fact doing when you submit a search query.

[1] https://support.jstor.org/hc/en-us/articles/115012261448-Sea...

[2] https://proquest.libguides.com/proquestplatform/tips