I recently setup an emulator, passed it through the TV and had blue tooth controllers hooked up, so I could show my son Donkey Kong Country - we ended up finishing 1 and 2 (with my wife joining in), he loved the older game. Trying to get him into older PC games now. So yeah, it is possible to get them into it (just dragging him away from minecraft is hard haha).
The updated Donkey Kong Country that is available on Switch is also a great game. From what I understand it is almost identical to the older version, but with updated graphics and the option to save progress
I have Aphantasia, as well as excellent map reading skills and spatial awareness. BUT I'm pretty sure that spatial awareness is something I developed when cheffing (needing to know where things were rapidly) and map reading I attribute to video games (specifically doom originally haha).
B) Warning everyone, it slowly eats tokens - If you're like me and on the pro plan, perhaps not a good idea to have its (albeit slow) drain on the tokens.
I set it up and saw the token notifier down the bottom slowly ticking up. I'll try again and report back, perhaps I misread something? I was surprised something like thst (which I assumed was an API call) was trickling token usage.
When was it ticking up? The status line only updates after each turn. If you submitted a turn, tokens were used. The status line itself is a bash script, you can go look at it.
(former chef here) - I've been writing my personal recipes like this for 20+ years. Mostly because I can grok the steps from memory so I can skip that part (other times I just write the spec step between the curly bracers).
Not the neatest, but my 'Dad's cookbook' isn't really read by anyone except myself when asking the family what they want me to prep on the weekend for the week.
I would argue the reason is so people can make arguments like yours.
Step 1: get easily passed, simple looking laws passed to 'protect the children'
Step 2: 'oh look at all these people bypassing the law. It's so simple for a child to watch porn with this.'
Step 3: increase the requirements bit by bit on the verification
By letting something simple pass, they can claim it's not thst bad, and anyone who argues against it is being hyperbolic.
> Step 2: 'oh look at all these people bypassing the law. It's so simple for a child to watch porn with this.'
This would be a demonstrably stupid argument. A parent who simply opens the box and does 5 minutes of setup before giving their child a new $300-2000 device, and maintains the most basic awareness of any rogue devices (i.e. ask questions if a new phone appears) is all it takes to make sure no "bypassing" takes place in your home.
Anyone claiming "all these kids are bypassing it" would be lying, unless the law isn't even being enforced or sites aren't cooperating. And no law can overcome those problems.
Do you have children? Have you witnessed what they can do with restricted $300-2000 devices? My kids hack restaurant tablets at every place we go. I don't even know how they do it most of the time.
This is going to be one hell of a slippery slope. And the actual publisher-driven proposal in this thread is so much more useful; I would love to filter out anything with excess violence and generally NSFW content from my view of the Internet -- and I am over 18.
Tinfoil hat, are they and anthropic doing this deliberately to increase the chance of the US govt licking down foreign models(well attempting, we all know how badly that will work).
I was doing something similar where I saved user input/model output in a multi-depth node style storage system (each depth having more precise details) with the focus on the model having accurate user fed information. I was mostly focused on retrieval of accurate / useful information based on user query (injecting the database node as additional, high confidence information)
Once this(Zero-mem) passes it's peer review, I may have to see if my system can handle something similar instead/in addition.
I'm quite excited to see growth in these different ways of eliminating token's.
Long winded aside, @langs, have you published your work on this?
Using attention for retrieval was inspired by a comment I saw in here long time ago: Prediction and retrieval are two sides of the same coin; to predict better, you must retrieve more accurately.
I'm still working on the improvement of algorithms, my tests shows the performance and accuracy will be improved a lot in the next release.
I am not an expert on LLMs, but what's preventing you from treating the context as 'virtual memory', and using the attention matrix to 'blank out' tokens which have very low weights and will not contribute much to the input? I imagine most tokens are like this, and you can skip computation on 95% of an 1M (or practically infinite) context.
"The value on the market [was] about three million dollars, but they would not let us sell it. And for that reason, we just [left] it there.”"
I think that's the point OP is pointing out, for better or worse (for what it's worth I believe "He suggests that a prior conflict with the CFIA over product labeling motivated them to close the factory before the inspection began" may be correct).
A company should have plans for cleaning up their product before just walking away and leaving other people to clean up their mess.
Maybe it's rather naive of me, but that's what I believe.
Yeah, Android's sandbox deals with low-level file access and socket APIs and stuff. But Android also, intentionally, allows apps to expose their data and functionality to other apps via the higher-level "intent" system.
Apps get to choose what permissions are needed to access their intents, so under Android's security model, really it's Chrome's fault (or whatever browser the user has installed) for exposing an intent that allows apps that don't have the INTERNET permission to exfoltrate data.
Similarly, apps are also allowed to collude to share data with each other if they want; that's how stuff like Google Play Services works.
> Apps get to choose what permissions are needed to access their intents, so under Android's security model, really it's Chrome's fault (or whatever browser the user has installed) for exposing an intent that allows apps that don't have the INTERNET permission to exfoltrate data.
You might as well blame the phone for not being encased in concrete and thrown into a well. You might not want your text message application to have the internet permission, but you'd certainly want to be able to open links from it.
As I understand it, this "Application Sandbox" is to prevent app A seeing what app B is doing.
The problem here is a different one: The user granting an app access to their files and then having no way to prevent that app from sending that data to the developer of the app.
