Hmmm… this is giving me thought actually. Given the choice between that and the current administration where the goals of self destruction are strongly in evidence, it’s actually worth thinking about. At least. Let me get back to you :)
On a tangential note, I’m curious if researchers have started running virtual simulations, where sandboxed AIs are used as decision makers of key political and business positions?
We also associate laboratories with evil scientists and Frankenstein and the like. I can just hear Boris Karloff (er Bobby Picket) uttering “I was working in the lab late one night. When my eyes beheld an eerie sight… … … …the monster mash”. If anything, I associate _uncertainty_ with labs. The result is never known up front, they’re a place of discovery.
But I get your meaning. What should they be called instead? AI Sausage Factories maybe (cue Upton Sinclair?)?
Even kind of fits the model. All of the creativity man has raised is herded to the slaughterhouse and ground up so we end up with a big homogenized mash of ground up creativity, devoid of the life that gave it, rotten if not eaten soon enough.
Everyone gets to name their projects for their own reasons. If the pitch is basically an “alternate apple” macOS, I would have chosen one of the other other (alternate) apple varieties as the root of my OS. Missed opportunity I guess.
Real apple-people would'nt consider this anyways. Open source geeks is were they can get traction. Seems like there is plenty of us that would like fbsd quality and vertical integration to the desktop level. There is effort but no breakthroughs.
The more technical side of the Mac user base would probably jump ship if a polished, open macOS analogue appeared. There's been a lot of discontent with changes dating as far back as 10.7 Lion, and has been gradually mounting more rapidly since 10.10 Yosemite and accelerating greatly with 13 Ventura through 26 Tahoe.
There's substantial demand for a spiritual successor to OS X 10.4/10.6/10.9 that isn't being filled.
I agree. I would switch to an open source spiritual successor to even Tiger in a heartbeat if it were available.
The difficult part is replicating the ecosystem of applications. From a technological standpoint I would prefer a modern Linux distribution with KDE or Cinnamon over using 2005-era Mac OS X Tiger, though I’d gladly use a version of Tiger that was retrofitted with a modern Darwin core, modern libraries for security purposes, and an up-to-date web browser. However, Linux does not have Microsoft Office, the Adobe Creative Suite, Preview.app, Keynote, Dictionary.app, and other pleasant Mac apps. The FOSS alternatives are not as polished. Even venerable Office 2008 for Mac feels nicer than LibreOffice, and it’s compatible with the XML-based Office formats.
I guess for now I could run Tiger in an emulator, though admittedly current operating systems serve my needs, even if I prefer the polish of Tiger and Snow Leopard. I get really good performance running PowerPC Tiger inside QEMU on my base M4 Mac Mini and on my Ryzen 9 3900 desktop, definitely not as snappy as running Tiger on actual contemporary PowerPC hardware, but fast enough to actually be productive.
I think this is one of those cases where “build it and they will come” applies.
The corporate giants are their own thing, but indie/small Mac devs are among some of the most steadfast early adopters in the software world. They routinely have new releases leveraging whatever new things were added at WWDC in the span of a few weeks or months. Many would be thrilled by a project that promises to bring back what they loved about OS X and would be happy to contribute.
The only problem is that the “it” that needs to be built is a set of frameworks that’s as robust as Cocoa/AppKit is, is flexible with language use (Mac apps have long rolled in C, C++, Ruby, and Python alongside Objective-C or Swift for UI), and isn’t afraid to be opinionated (a single well-supported “happy path” to do something is better than a dozen halfassed ways). That’s really the hardest part.
As for Tiger, on the Ryzen box you might get better performance from a hackintoshed x86 build (especially if you’ve got a spare PCI slot for passing through an old supported GPU for UI acceleration).
Tell me about it. Those OS X versions you mentioned for Tiger, Snow Leopard, and Mavericks were some of the best releases of OS X ever.
Thankfully, macOS Golden Gate is proving to be a decent release so far in my tests of the public betas, but I do yearn for the simplicity (and more power user friendly aspects) of the aforementioned OS X releases.
As long as message broadcasting remains cheap (in any form; phone, email, text, etc) it will go on. All modern media has evolved technically (and financially) to benefit the producer (the talker). All at the expense of the consumer (listener). Make outgoing bandwidth cost per instance and you’ll see a reduction/optimization of the production. When signal is free to produce, you end up with a lot of noise.
I would love to see what this looks like in practice. Especially in generated code. I assume this is more than insertion of non visible special unicode whitespace characters, but more in the pattern of the text content itself?
Models can't reliably follow instructions involving their own logprobs unless they can take agentic control and use quite sophisticated dynamic grammars/structures/constraints to force this behavior in one shot (which can be slow and the dynamic grammar modification feature isn't supported in closed model APIs for safety reasons) or repeated attempts at rewriting which is expensive/slow.
Yes they can do this, but it's more likely closer to the original "red token, green token" paper: https://arxiv.org/abs/2301.10226
i.e. take half of your LLMs vocabulary, and upweight its probabilities by ~55% to the other half's ~45%, and scan for overuse of this half of all tokens. You can even choose a different half/slice for every individual user, for every individual action. You can implement this under the hood cheaply with logit-biasing.
Well put. I wonder if this will inevitably increase usage costs because the model has to do more work to find a solution that meets watermarking standards.
Considering how weirdly detuned tokens selections have become in Anthropic's LLM prose in recent models, there is a chance this goes unnoticed in everyday use.
IIRC, watermarking text could be as simple as training the model to use specific words/phrases more frequently than what you would expect to find in human-written text, to the point where it's highly statistically improbable that it wasn't AI generated. I assume similar logic could apply to code in the form of functions/code styling.
That's probably an over simplification. Also a solid defence that can be used against complaints about the way AI writes text.
It essentially looks like the difference between two different runs of the model with the same prompt but different seeds. The watermark is essentially a small bias in the model such that when there's multiple different tokens that could conceivably follow the previous token, the model will only pick some subset of them (the subset is derived from a hash of the previous token). This bias can then be checked for statistically (without needing access to the model and without needing the whole prompt), and for longer text where there's enough freedom in word choice you can show that it would be vanishingly improbable to accidentally follow the rules in the watermark.
Your comment kind of reminds me of the old slashdot spam-email-solution copypasta - the purpose of it is to highlight how so many sensible sounding ideas just can't work in practice.
Honest ask because I’ve not generally paid attention…
Is it the case that the numbers always appear more optimistic, and then are revised to be more pessimistic values? Or is it just that the original numbers are off and are revised? Does that make sense?
> They move in both directions regularly, typically in the direction the economic situation is moving
And for good reason! These are surveys. The easiest respondents are those saying “no change.” The latest responses, the ones who had the biggest change. In a booming economy, the biggest changes will be hires. In a troubled one, the biggest changes will be those doing lay-offs.
Add to that: the BLS tries to correct for this effect. Which means at the moment the economy turns, its corrections skew the wrong way. (On the whole, per the above chart, the figures are getting better over time.)
I totally see this. I just did 3 hours of bot sitting to put together some thrash loops that thrash our provisioning working flow for a BLE gadget we make. It was pretty straightforward and productive. But then, I have a lot of experience with BLE, and a quite a bit of experience with python and shell scripting. So I was able to guide the process through stages, do some intermediate testing, make some adjustments, and proceed. Domain experience made this really easy and straightforward. Me two junior engineers who have only superficial/high level knowledge of BLE and some of the other pieces, couldn't have done this as effectively.
Where my angst comes, is worrying that no one will ever get that experience anymore. They might have had some eventual success, who knows what monstrosity a much less guided LLM would have done, but experential learning may be mostly a thing of the past. And it creates a real tension between the person with experience and the person without.
>Where my angst comes, is worrying that no one will ever get that experience anymore.
I am a fairly inexperienced python developer using LLMs to build software and find that I still learn a lot just from prompting and tinkering. Maybe that's less true once you reach a certain level of competence, but at my intermediate level I am still learning a lot even leaning heavily on LLMs.
Hmmm… this is giving me thought actually. Given the choice between that and the current administration where the goals of self destruction are strongly in evidence, it’s actually worth thinking about. At least. Let me get back to you :)
On a tangential note, I’m curious if researchers have started running virtual simulations, where sandboxed AIs are used as decision makers of key political and business positions?
reply