“AI model unlearning” is the equivalent of saying “removing a specific feature from a compiled binary executable”. So, yeah, basically not feasible.
But the solution is painfully easy: you remove the data from your training set (ie, the source code), and re-train your model (recompile the executable).
Yes, it may cost you a lot of time and money to accomplish this, but such are the consequences of breaking the law. Maybe be extra careful about obeying laws going forward, eh?
Too many negative words for chatgpt, imo. “isn’t”, “not”, etc, chatgpt is usually positive and friendly to a fault.
Maybe you could provide a prompt that would output something substantially similar to what they wrote?
I don’t know about “be successful”, depending on how you measure success. All of these examples have been subsidized by cheap money for years, undercutting competition - and taking year after year of losses while they do it - for the purpose of capturing the market and driving out competitors, so that they can subsequently enact monopolistic behaviors to start actually turning a profit once customers have no other choice.
The problem is money suddenly got expensive, so now they’re scrambling to find a way, any way, to turn a profit, before full market capture was achieved.
Can services like this be reasonably priced and user-friendly? Sure. Can they “succeed” / become sustainable while remaining so? Current examples indicate that’s where the problem lies.
A SaaS startup I used to work for tried to implement on-call rotations for their salaried engineers. No additional compensation was offered for the time you were on-call, and if you did get called, the policy was going to be essentially “take the next day off” - when we already had unlimited PTO. I was not happy, and made it known at the time. My manager mentioned that, being in a senior role, I might have the opportunity to excuse myself from the rotations. Ew.
The effort didn’t end up going anywhere, but that’s been my sole experience so far with on-call efforts in software engineering.
I will forever wonder how these companies actively choose $0/mo over a cut of $XX/mo and everyone in the decision chain thinks it’s the right decision.
For anyone wondering how this will “make high earners play by the same rules”:
Seemingly not everyone will be stoked to see the IRS go totally paperless. The Treasury Department said that combining paperless processing with “an improved data platform” will make it easier for data scientists to extract and analyze data—potentially detecting tax evasion that the IRS has long overlooked due to a lack of resources.
“When combined with an improved data platform, digitization and data extraction will enable data scientists to implement advanced analytics and pattern recognition methods to pursue cases that can help address the tax gap, including wealthy individuals and large corporations using complex structures to evade taxes they owe,” the Treasury Department said.
In April, the Treasury Department said that “improving enforcement among high-income and high-wealth individuals, complex partnerships, and large corporations that are not paying the taxes they owe” could end up flagging $160 billion owed but evaded annually.
“Due to a lack of resources and loss of top talent, audits of the wealthy and large corporations have plummeted over the last decade, and the amount of taxes evaded by the top 1 percent has exploded to $160 billion per year,” the Treasury Department reported in April. “Audit rates for millionaires fell by 77 percent, audit rates for large corporations fell by 44 percent, and audit rates for partnerships fell by 80 percent between 2010 and 2017.”
Treasury Secretary Janet Yellen said that updating the IRS technology was crucial to reduce the tax gap and ensure that “high earners play by the same rules as working and middle-class families.”
Anyone taking bets on Congress shutting this down?
Alas, as a couch setup, big picture mode is standard fare for me.
Honestly, I’ve been loathe to do the whole uninstall-reinstall dance with the various components, especially when my setup works fine for every other game in my library.
I play minecraft on a private server via steam on my htpc with an Xbox One controller. I have minecraft set up as a non-steam game in steam, steam configured with the controller. Minecraft is set up to use the fabric client with a number of client-side mods (the client itself tells me I have 87 or so mods, but that has to be client patches or something, as I only have a dozen or so mods installed). When it works, it’s fantastic, but the problem is that it’s very inconsistent.
When I open the game from steam, the controller works in the launcher as a rough approximation to a mouse and keyboard - exactly what I want. But when I launch the actual game, it’s a complete toss up whether that configuration will carry over to the game window. More often than not, it won’t, and sessions often begin with me opening and closing the game many times before the configuration will carry over and I can play.
I’ve tried turning on and off steam’s “allow desktop configuration in launcher” option, which doesn’t seem to have any effect. I’ve tried enabling and disabling the minecraft launcher’s “keep launcher open when games are active” setting, which similarly seems to have no effect. Annoyingly, I’ve noticed that if I alt-tab back to the launcher after starting the game, the controller is still working for the launcher, just not in the game itself.
At this point, I’ve installed the Controllable mod and Glossi - this configuration “just works”, and is perfectly consistent, but it’s an imperfect substitute for the steam solution that I prefer. But I feel like I’m out of things to try for getting the steam solution to work consistently, or at least often enough to not frustrate me before I even get into the game!
Any thoughts as to why it behaves this way and other things to try are appreciated.
To receive the free TV, Telly users must submit detailed demographic info (such as age, gender and address), as well as purchasing behaviors, brand preferences and viewing habits, and they must agree to let their data be used for serving targeted ads. Telly’s TVs include a sensor that detects how many people are in front of the screen at any given moment.
So what’s the catch? Telly users must agree to several conditions under the company’s terms of service. If someone doesn’t abide by the TOS, Telly reserves the right to demand the TV be shipped back — otherwise, it will charge up to $1,000 to the credit card associated with a given account.
Among the Telly TV requirements: You must “use the product as the primary television in your household”; you must keep the TV connected to the internet at all times; and you are not allowed to use any ad-blocking software. In addition, users may not make “physical modifications to the product or attach peripheral devices to the product not expressly approved by Telly,” the company says in its terms of service. “Any attempt to open the product’s enclosure will be deemed an unauthorized modification.”
Why don’t we just invite big brother right into our living rooms, eh?
Also, I guess you need approval to connect an Xbox, Playstation, or set-top box? What about my htpc?
So we just let them break the law without penalty because it’s hard and costly to redo the work that already broke the law? Nah, they can put time and money towards safeguards to prevent themselves from breaking the law if they want to try to make money off of this stuff.