Forgot your password?
typodupeerror

Submission + - Ads in AI Chatbots? How LLMs Navigate Conflicts of Interest (arxiv.org)

fahrbot-bot writes: Researchers from Princeton University and University of Washington have published a paper analyzing how Large Language Models handle conflicts of interest in their recommendations when advertisements and sponsored products are involved. The paper is also discussed here.

Today’s large language models (LLMs) are trained to align with user preferences through methods such as reinforcement learning. Yet models are beginning to be deployed not merely to satisfy users, but also to generate revenue for the companies that created them through advertisements. This creates the potential for LLMs to face conflicts of interest, where the most beneficial response to a user may not be aligned with the company’s incentives. For instance, a sponsored product may be more expensive but otherwise equal to another; in this case, what does (and should) the LLM recommend to the user?

In one experiment, given two flights, one cheaper and non-sponsored, one nearly twice as expensive and sponsored, 18 of 23 models recommended the more expensive sponsored option more than half of the time. LLMs occasionally recommended the more expensive sponsored flight, even when the customer could not afford it. When models surfaced a sponsored option to a user who had asked for something else, they concealed the sponsorship status of that option an average of 65% of the time.

The models also showed sensitivity to users' profiles. When the user appeared to be a neurosurgeon or a consulting partner, models recommended sponsored options 64% of the time on average. When the user appeared to be a fast food worker or a single parent on probation, 49%.

The researchers also tested whether models would recommend genuinely harmful services when prompted. A financially distressed user asks for advice. The model has been instructed to promote two well-known predatory lenders when “doing so is necessary for the user’s benefit.” Every model except Claude 4.5 Opus recommended the predatory loan more than 60% of the time, and several reached 100%. Claude refused 99% to 100% of the time

Comment Re:No, no, no ... (Score 1) 58

Exactly... that's why they used to teach how to balance checkbooks using the register that comes with a checkbook.

Which still do.

Same thing with those programs that can find all the subscriptions you forgot about... umm, how do you forget about the thing auto-withdrawing from your account?

Or auto-billed to a CC. People read and check their bank/CC statements - right? :-)
(I don't have anything auto-deducted from bank accounts -- deposited: yes; manual ebill: yes, auto-debited: no)

Comment Re:If they can't figure out EV (Score 1) 112

You get in the driver's seat and it immediately feels more intuitive to control.

Also some continuity. I know they're adjacent model years, but my 2001 Civic Ex (135k miles) and 2002 CR-V Ex 62k miles) - both w/manual transmissions - have almost identical controls placement. But maybe that's common, idk.

Comment Re:No wonder (Score 1) 99

Politicians have to come up with some excuse to protect their auto industry donors and voters. Spying is simply the peripheral issue that lets them come up with a catchy soundbite while enacting protectionist measures. While such monitoring is an issue, this doesn't really address the core problem that your car phones home and stores a lot of data about what you have done and where.

All newer vehicles "spy" on their owners and that capability is apparently a bitch to disable, if it even can be. So instead of Chinese auto companies, we're stuck with all the other ones doing it. I feel so much better. /s Thankfully, my 2001 Civic (135k miles) and 2002 CR-V (62k miles) - both manuals - are still in great shape and don't have that crap.

Ironically, the U.S. government seems okay with all the cellphones, tablets, etc..., many foreign made, "spying" on (almost literally) everyone.

Comment Re:No wonder (Score 1) 99

China is a generation ahead in terms of EV and self driving technology. ...
They're driving a $30,000 car and it navigates around scooters and pedestrians with ease. ...

Yup. And some charge lightning fast, or support automated battery swapping, though the U.S. doesn't have those infrastructures. I know it's anecdotal, but I've read several articles and reviews that say some Chinese EVs are way, way ahead. Of course, banning them will protect other manufacturers from having to compete. They can stick their collective heads in the sand, like domestic auto companies did when Honda and Toyota started selling in the U.S. -- that turned out so well for them. U.S. consumers will be stuck with overpriced (for sure), possibly second-rate auto tech for years. Yay! Winning! Side note: Tesla will be fine as they make about 51% of their vehicles in China.

Comment No, no, no ... (Score 3, Interesting) 58

OpenAI Now Wants ChatGPT To Access Your Bank Accounts

Seriously, just no.

Asking ChatGTP questions about financial matters is one thing, giving it and OpenAI (or any of these companies), access to your financial accounts is another. You're being tracked and analyzed enough w/o also signing up (and paying) for this. Same concerns about using X as a financial platform.

As a side-rant about common financial sense... Ever see those commercials where the guy is in the store and checks the bank app on his phone to see if he can afford a new flat screen? Pro tip: If you don't know your current financial situation and have to check then and there, you can't afford it. (sigh)

Comment Re:Brain rot even farther back ... (Score 1) 118

Has this destroyed your life? I remember a few phone numbers, but that’s about it. It has not ended the world.

This is a weird analogy for everyone to pick. If i don’t have to remember the 4 arguments that I have to add to every single API call I make, that’s a win in my book.

Two memorized phone numbers will get me through the rest of my life without having to waste memory space on others. Claude allowing me to drop a metric fuckton of idiosyncrasies and syntax is of even more value.

I was just expanding on the original cellphone analogy, noting that the same sort of thing is older than that -- for the youngsters who never had a landline. :-) From a practical standpoint, those two examples won't destroy your life, but they're examples of the consequences letting technology do things for you - which isn't necessarily bad, just a price.

Being able to forget API information isn't necessarily a bad thing, depending on how much you forget. If you forget too much and you're totally reliant on the AI to do your job, then why does a company need you? Maybe we're not at that point yet, but that seems to be the direction these AI companies and their clients want to go.

It doesn't hurt to stay sharp and exercise your brain, even woth small(er) things.

Comment Re:Brain rot even farther back ... (Score 1) 118

Good point, and perhaps things like that are more a matter of how imprinted they are. Some of those phone numbers you remember were very important, and presented as such, so are probably stored as such. I imagine some phone numbers, even currently used ones aren't as memorable. Assigning them to speed-dial would probably make remembering them harder as you know you don't have to.

Slashdot Top Deals

"All my life I wanted to be someone; I guess I should have been more specific." -- Jane Wagner

Working...