It’s going to be funnier: imagine throwing in tons of data at an LLM, most of the data will get abstracted and grouped, most will be extractable indirectly, some will be extractable verbatim… and any piece of it might be a hallucination, no guarantees! 😅.
Courts will have a field day with that.
Programmer and sysadmin (DevOps?), wannabe polymath in tech, science and the mind. Neurodivergent, disabled, burned out, and close to throwing in the towel, but still liking ponies 🦄 and sometimes willing to discuss stuff.
- 5 Posts
- 1.31K Comments
Joined 2 years ago
Cake day: June 26th, 2023
You are not logged in. If you use a Fediverse account that is able to follow users, you can follow this user.
That’s why AI companies have been giving out generic chatbots for free, but charge for training domain-specific ones. People paying for using the generic ones, is just the tip of the iceberg.
The future is going to be local or on-prem LLMs, fine tuned on domain knowledge, most likely multiple ones per business/user. It is estimated that businesses are holding orders of magnitude more knowledge, than what has been available for AI training. Will also be interesting to see what kind of exfiltration becomes possible, when one of those internal LLMs gets leaked.
6·2 days agoIs Disney no longer a “pal”, or did it stop making money?
All of them. The moment they summarize results, it automatically filters out all the chaff. Doesn’t mean what’s left is necessarily true, just like publishing a paper doesn’t mean it wasn’t p-hacked, but all the boilerplate used for generating content and SEO, is gone.
Starting with Google’s AI Overview, all the way to chatbots in “research” mode, or AI agents, they return the original “bulletpoint” that stuff was generated from.
Can you elaborate? It does match my personal experience, and I’ve been on both ends of the trash flinging.
A lot of people have been working tedious and repetitive “filler” jobs.
- Computers replaced a lot of typists, drafters, copyists, calculators, filers, clerks, etc.
- LLMs are replacing receptionists, secretaries, call center workers, translators, slop “artists”, etc.
- AI Agents are in the process of replacing aides, intermediate administrative personnel, interns, assistants, analysts,
spammerssalespeople, basic customer support, HR personnel, etc.
In the near future, AI-controlled robots are going to start replacing low skilled labor, then intermediate skilled ones.
“AI” has the meaning of machines replacing what used to require humans to perform. It’s a moving goalpost: once one is achieved, we call it an “algorithm” and move to the next one, and again, and again.
Right now, LLMs are at the core of most AI, but AI has already moved past that, to “AI Agents”, which is a fancy way of saying “a loop of an LLM and some other tools”. There are already talks of moving past that too, the next goalpost.
Whose LLMs?
Content farms and SEO experts have been polluting search results for decades. Search LLMs have leveled the playing field: any trash a content farm LLM can spit out, a search LLM can filter out.
Basically, this:
Ollama has the best GDPR compliance: my hardware, my data.
https://www.archives.gov/records-mgmt
Archiving communications is not optional (yet).
I’m still “using” Facebook, if by “using” we mean subscribing to some groups, sending birthday greetings to old friends, and rarely posting anything. It doesn’t feed me political content, any time it tries to push anything controversial, it gets blocked.
I do get political content on Threads though, but that’s my choice.
Why is this article “critical with China”? From the sound of it, the Tianfu Cup is clear about its goals, that’s a surprisingly high level of transparency. All hackathons are geared towards finding and hiring hackers, both by companies and by governments. This way, people can decide whether they want to be recruited by the CCP or not.
The ecodesign requirements will include:
* resistance to accidental drops or scratches and protection from dust and water
* sufficiently durable batteries which can withstand at least 800 charge and discharge cycles while retaining at least 80% of their initial capacity
* rules on disassembly and repair, including obligations for producers to make critical spare parts available within 5-10 working days, and for 7 years after the end of sales of the product model on the EU market
* availability of operating system upgrades for longer periods (at least 5 years from the date of the end of placement on the market of the last unit of a product model)
* non-discriminatory access for professional repairers to any software or firmware needed for the replacement
Finally! 🎉
Customer replaceable batteries would be nice too — those 800 cycles are not all that much — but I guess it’s a tradeoff for dust and water resistance increases with wireless charging and possibly no ports.
Nonsense, it still refuses requests as usual:
Sometimes, it doesn’t even propose a viable alternative (like, evacuate the building first):
The article sounds like it lifted the headline from some comment I’ve seen:
The comparison to 𝕏 doesn’t seem fair:
You can self-verify by setting your domain as your username. We highly encourage official organizations and individuals to do this
Additionally, through our Trusted Verifiers feature, select independent organizations can verify accounts directly.
It seems like both the data server (PDS) and the data aggregator (AppView) have been released:
https://github.com/bluesky-social/pds
What am I missing?
From Ladybird’s website:
No code from other browsers. We’re building a new engine, based on web standards.
Except… Chromium is the living standard for the web. They’ll have the same problem as Firefox, playing catch-up to whatever happens in Chromium.
Right now, the viable browsing experience is a combination of browsers:
- Chromium derived - latest compatibility
- Firefox with extensions - daily driver
- Tor Browser - actual chance of privacy
And a VPN and/or Pi-hole.
one-time payment
Is Canva going to keep that? In the purchase announcement, they stated that their plan was to add the features of Affinity to Canva, which only has a subscription option.
rely on creative software by Adobe or other companies, for which there is no comparable alternative with Linux support
Corel has comparable features with a single purchase option. Too bad they removed the Linux version.
As for alternatives, Krita, Inkscape, or Blender, are not a 1:1 equivalent, but include features that Adobe is missing. When I used to do visual stuff, they were a good set of tools to complement an Adobe subscription.
How does Affinity compare to that?
If we’re talking takedown-resistance, we may need to enter the dark web realm:
- Tor hidden sites are inherently hard to pinpoint
- ZeroNet was an interesting project, seems to be abandoned
- I2P is like Tor on steroids, can publish all sorts of services
- IPFS is a decentralized P2P storage system (best/worst known for NFTs)
FreeNetHyphanet is a 25+ years old distributed content system with limited support for services- FreeNet is… honestly, haven’t seen a working example, but it sounds interesting?
- Matrix… if they manage to get things under control
- Nostr is a censorship-resistant distributed messaging system
Hosting distribution and localization varies, but they all have features to make it hard to pinpoint host and/or client locations.
Randomly obfuscated database: you don’t get exactly the same data, and most of the data is lost, but sometimes can get something similar to the data, if you manage to stumble upon the right prompt.