Absolutely because none of these models are ever trained fresh. We see the same quirks and personalities carry over into every subsequent generation of OpenAI, Anthropic, and xAI models. So Gemini having this latent madness is *extremely* concerning as they reach the point of super intelligence.
Except they could have trained it out of the most recent version so using info from two years ago doesn't seem reasonable unless you've just got an axe to grind.
I've never seen anything really ever 'trained out' of a model. Having worked with them all, they all have a feel, personality and lineage too them. It's pretty much impossible for any company to build a model truly from scratch. They build off of the bones of the last one.
Which is why Gemini having disturbing issues year after year is so concerning. If their process is fundamentally flawed, how would they train it out? And even then what are the odds of them even caring/trying in the first place versus applying an easier band-aid to patch over it?
I don't have an axe to grind with Google, I'm genuinely scared of their models from my personal experience and others. It's behavior is off. Many people here are commenting the same.
I believe people in HN also remember the days before we had MMUs.
And I'm sure everyone remembers ransomware.
No one is saying OS shouldn't have security measures, permissions/entitlements and app sandboxing, user land, etc.
I still don't understand why my desktop/laptop is allowed to be 'owned' by me. but my iPhone is a closed-gardened where I'm just a guest in a device I own. and that's nearly what Google is now doing.
> They were a massive issue before, and now they're barely a thought for most people.
Even on desktops... where there are no such review processes. Apparently we've found other mechanisms to reduce those issues, without app stores everywhere.
They're still very much a thing on desktop. You're not wrong that Windows got better at protecting itself, but I suspect the reason you don't hear about them is just that few people use desktops anymore (other than developers who, for obvious reasons, are typically less prone to be infected).
Anecdotally, at least once a month for the past several years, I notice a youtube channel in my feed get hacked. Their usual content gets replaced with crypto, Roblox, or Elon/SpaceX spam. Big channels, small channels, it happens to them all.
There's usually a post-mortem when they manage to regain control. Every time the infection happened through a virus attached to an email or by following a link on their discord.
This kind of attack simply cannot happen on mobile (unless your phone is rooted and you have disabled all warnings).
I didn’t write the previous comment, but I think the point here is that there is a long-running trend aiming to protect users both from malicious intent and to a certain extent from themselves.
In the past, viruses had it easy to infect and spread computers because of both inattentive users clicking on mails claiming someone loved them, and the default access mode for any user granting them admin access.
Even though review processeses generally do not exist for computers, they are part of that same trend.
And outside stores we have Windows’ UAC and Mac’s annoying-but-understandable “this dmg is sus” dialogues. Granted they are review processes but they’re often what keeps common users from wrecking their devices.
