> Which brings it much closer to LLMs in this case (can an LLM talk about a topic it has never seen in training? unlikely if it cannot be derived from other training data)
This is a great experiment. Do some world building offline about Smergle the Keen finding the Hoopla of Kynashiusous. Then ask ChatGPT what reagents would have been more effective in breaking through the borantiastic door than squeeble stomach juices.
Of course we know the answer - it will either drop into a 'I have only been trained on data before blah blah blah' or 'I cannot talk about current world leaders' or 'I have not been trained on your hackneyed and unpublished romantasy saga, why don't you tell me about it?'
On the other hand, if it is aware of the general concepts of your concept, and has bridging concepts for related topics it has been trained upon, it might try to apply the general concepts to a specific but different scenario.
AI has useful features but the majority of use cases are not useful but are thrust upon us as an excuse to extract more data.
Even if the worst happens in the AI market some uses will still remain. Far fewer, I suspect, than will be demand the computational capacity of all the DCs popping up everywhere.
The goal of LLM's, as they are marketed now, is to drive engagement and stickyness of products. A wrong answer is brushed off with a "Hey, you're right, let's try that again" - a response purposely designed to maximise the friendliness of the system and minimise the sting of a wrong answer. The fact that an LLM will not respond to the same question in the same way twice (i.e. the 'temperature' ) is because increased accuracy will not drive engagement and, therefore, increased accuracy cannot be allowed to get in the way of engagement.
Glad someone else felt the same. Discovery started well but the whole "burn" thing was... stupid. There are a million narrative devices you could use to break the ability for ships to travel via warp and they chose the least sciency of them all.
Remember, the Cerner acquisition enabled an essentially permanent beachhead for Oracle in the VA. It will pay off over time in 'we already run Oracle' across many dimensions of government.
They had a very kludgy, homegrown, but generally well-liked and cost effective EMR calls Vista.
The fact that it was cost effective, but required internal expertise was exactly what led the Trump administration to kill it.
Now they are transitioning to Cerner, the worse of the two major commerical EMRs by a fair margin, for many billions of dollars to the Kansas-based company, now owned by Oracle.
You can fill in the details of why this is happening by looking at where the CMS appointees at the time have landed in their post-government careers.
> Perhaps aesthetic - both Windows 1.0 and 2.0 were (to me at least) very ugly.
But it was amazing for those of us used to black and white/green/amber screens in DOS. You could put an image as your background. And it stayed there, lurking behind your word processor or spreadsheet, to spring back into your vision whenever you finished up your work.
One of the very first things I always do in any OS is to set the desktop background to solid color, usually black. I almost never ever will see it, coz there's always going to be a window on top of it, except upon startup for some brief period of time or if I accidentally minimize everything.
I work with full screen windows. Always (tiny number of exceptional cases maybe). I switch between windows with Alt+Tab when necessary. I also have a relatively small screen, both for work and personal stuff (14" for at least 10 years now).
I was doing tech support through the Windows 95/98/ME period and it was hell. Everything either crashed the OS or required a restart if you touched it.
When Windows 2000 rolled around and I saw how stable it was, I went out and bought it to put on my gaming PC. Another friend from work laughed at me and told me how terrible "Windows NT" was for running games until he saw how smooth Starcraft ran on it.
This is a great experiment. Do some world building offline about Smergle the Keen finding the Hoopla of Kynashiusous. Then ask ChatGPT what reagents would have been more effective in breaking through the borantiastic door than squeeble stomach juices.
Of course we know the answer - it will either drop into a 'I have only been trained on data before blah blah blah' or 'I cannot talk about current world leaders' or 'I have not been trained on your hackneyed and unpublished romantasy saga, why don't you tell me about it?'
On the other hand, if it is aware of the general concepts of your concept, and has bridging concepts for related topics it has been trained upon, it might try to apply the general concepts to a specific but different scenario.