As an IT security loser it’s not your software I’m worried about, it’s your OpSec. I’ve seen enough innocuous software being owned and being used to deploy malware that it takes a lot of due diligence to trust providers of random binaries
We have a process at work where clients export information from their database as a pdf which they email to us so that we can ocr it and insert into our database.
No one else seems to think this is bat shit insane
The Omega Factor is another good one you’d like judging by the rest of your list.
Was terrified by the theme music as a child and never allowed to watch it. Managed to find it a little while ago and loved it. Only one season unfortunately.
I’ve been rewatching it recently and thought the first season was awful. The plots specifically, the aesthetics and theme music are awesome and worth watching for that alone. Haven’t gotten to the second season yet but was really hoping it improved…
Biggest disappointment was reading up on it and finding that the Anderson’s had originally intended to do a second season of UFO then changed it up to make Space 1999 instead.
From memory I believe it was windows 7 that broke the taskbar - before that you could put it on any side of the screen and also embed folders on it.
I used to have it on the left side of the screen with my actual documents folder (not “My Documents” which even then was full of other crap) embedded in it. Kind of like vertical tabs in the browser, but better
I was really annoyed when they took that functionality away for no apparent reason. Win 7 was the start of the slide for me, windows steadily got worse with them removing more and more functionality. Was it seven that removed all the customisation you used to be able to do as well and replaced it with much more limited themes?
This isn’t a bug, it is known behaviour that is inherent and fundamental to the way LLMs function.
All the AI companies are aware of this and are pressing ahead anyway - it is completely irresponsible.
If you haven’t come across it before, check out Simon Willisons “lethal trifecta” concept which neatly sums up the issue and explains why there is no way to use these things safely for many of the things that they would be most useful for
I think you’re absolutely right. These companies know full well that their “guardrails” are ineffective but they just don’t care because they’ve sunk so much money into AI that they are desperate to pretend that everything’s fine and their investments were worthwhile.
I was on a call with Microsoft the other day when (after being pushed) they said they had guardrails in place “to block prompt injection” and linked to an article which said “_help_ block prompt injection”. The careful wording is deliberate I’m sure.
I really think we should stop using the term ‘guard rails’ as it implies a level of control that really doesn’t exist.
These things are polite suggestions at best and it’s very misleading to people that do not understand the technology - I’ve got business people saying that using LLMs to process sensitive data is fine because there are “guardrails” in place - we need to make it clear that these kinds of vulnerabilities are inherent in the way gen AI works and you can’t get round that by asking them nicely
Some languages - human languages, natural or constructed, not programming languages - have ways of expressing the source of a piece of information, such as whether it is a known fact, something you have personally observed, or just an idea you are throwing out there for instance, using grammar or extra signifying words.
I’ve often thought about stealing some of these techniques for note taking as it would be really useful to have extra context without having to write it all out longhand.
Check out lojban attitudinals for one example along these lines if you are interested. There are natural languages that do this as well but I can’t remember any specifics off the top of my head