tonyarkles

> While I don't know if AI is conscious, I'm sure it has an internal world. Isn't that what a latent space literally is? LLMs operate in a high-dimensional space that can be analysed and translated to very abstract, high level concepts, behavioural tendencies, even things recognisable as emotions.

From your comment:

> My wife says please and thank you to Alexa. She gives it consideration like she would another human. When the robot asks me what my plans are for the rest of the day, I picture a Jira epic sliding from “In Progress” to “Done,” followed by a confetti emoji from a product manager at Amazon trying to earn their nubby little devil wings.

Where things get super interesting to me is the fact that it seems like (I have maybe read some academic work related to this, not sure) the kind of politeness your wife exhibits isn't useless in the context of a multi-turn session. To the other commenter's point, my understanding is that politeness/rudeness exists in the training corpus (scraped from stackoverflow, reddit, forums, etc) and that you can end up in different areas of latent space based on the overall tone of the session.

In some of the chat tools, because of the context management and "personality" stuff that happens behind the scenes, this can extend cross-session as well. ChatGPT, for example, for me, will frequently include text like "Hell Yeah!" and "Here's the shitty part..." in the responses, which is something I assume is happening because of historical context that's getting injected at the start of each session.

All of that to say... none of this, to me, has much to do with whether or not the thing is conscious, but rather that the training data makes it incredibly difficult for people to not see them as emotional beings because you can end up walking into different areas of latent space where different "emotional states" are affecting the output you're getting.

Unlike your polar bear example, an LLM can "love" you or "hate" you based on the input tokens that lead to the generation of the output tokens. Which is a pretty wild result on its own. Playing around with, say, Qwen 3.8 27B locally with my own experimental harness has been exceptionally fascinating.

I’m not sure that integrating it into a DAW is the way to go, but there’s definitely a need there.

I appreciate the sentiment but I feel like there is a lack of creativity and imagination behind that statement.

Yeah, that’s one of the places where it gets really complicated. There was that story out of… Australia, I think, where someone asked OpenClaw to get them a slot in a morning gym class and the LLM figured out an unauthenticated API call it could make to cancel other peoples’ registrations to free up slots in the class. Very likely that that violated Australian law, even though nothing was “hacked” per se.

It's been really nice to be able to breathe new life into some old hardware I had kicking around that's been discontinued. Some of it just needed old exploits applied to it, some of it needed some binary reverse engineering. Lots of security tools are dual-use as well; a model that knows nothing about hacking is going to have a really hard to time helping defend a system that's getting hacked (see the HuggingFace incident where HF staff were unable to use ChatGPT or Claude to analyze the logs from the attack because they hit security research guardrails).

> manufacturing viruses at a scale...

I'm not sure if you're referring to software viruses or biological viruses here.

If you mean software viruses, see above.

If you mean biological viruses, a lot of this gets back into the dual-use nature as well. I brew beer and mead. This involves cultivating specific strains of bacteria and providing them with a medium where they can convert sugar into CO2 and Ethanol. I don't think there's a good way to thin-slice the training data so that it can be an expert helper at Saccharomyces cerevisiae cultivation while being completely naive to Clostridium botulinum cultivation. Even further, it seems that helping someone make sure that they're not inadvertently cultivating Botulinum (e.g. water bath canning with insufficient pH) is useful.

Genetic lottery for sure. I’m still quitting but it’s not really because of acute or chronic health problems, mostly cost and being tired of stinking all the time.

Step by step, I feel like it might actually stick for real this time. Every time I started up again in the past was contextual (eg a break up, out for drinks with friends, have a half drunk smoke and I’m hooked again). No longer have any smoker friends, no longer have significant drama in my life, hoping it’ll work out.

- The first board, which I had already done manual schematic capture, floor planning, and some routing on (but then I lost steam and it’s been sitting at 90% for 6 months) was finished quite readily by Opus + KiCadRoutingTools.

- A second significantly more complex board that has been sitting at the “I’ve done a bunch of planning in Obsidian but haven’t started schematic capture yet” stage… Fable took my Markdown notes, asked questions to fill in the blanks, and put together a plan. Opus first put together a very ugly schematic (essentially all of the parts but no wires between them, just net connections) which then got refined into a nicer-ish schematic. Floor planning went smoothly, the first cut was decent and with a little bit of manual refinement it got somewhere pretty good. Currently routing and still debating a bit between 4-layer and 6-layer. The BGA escapes seem to be causing the majority of the problems right now; originally Opus decided that 4-layer wasn’t feasible due to not being able to even route all the nets, switched to 6-layer, did better but still ended up with like 70 unroutable nets. A bit of back and forth led to revising the BGA fan-out and now we’re back at looking at whether a 4-layer board might be feasible.

For the second one, I’m not 100% sure whether wall-clock-time this is a faster process than it would’ve been to do manually, but… the initial 4-layer attempts were all done while I was out for a date with my wife, the remaining 4-layer attempts ran overnight, the first 6-layer routing attempt was done while I was having my coffee this morning, and now the 4-layer/6-layer fanout changes are being done while I run errands. Something I find especially interesting is that by leaning on KiCadRoutingTools, it looks like there’s actually very little token usage at the routing stage; the plan gets put together as Python + KiCadRoutingTools instructions and then the hard work is done by that tooling. LLM picks up the results, looks at it, tweaks, kicks off a new job. That’s a pretty interesting result; this has all been running on my laptop pretty much continuously for 15 hours and I’m nowhere near any of my usage limits.

I'm actually pretty curious about this. You can't easily sneak on/off a commercial international airplane, but that's because they've got the gates set up in the airport to funnel you through a specific path that leads to customs. Showing up on an international charter or GA flight seems squishier. http://www.canada.eaachapter.org/FLYING%20TO%20and%20FROM%20... The process seems to be that you schedule an ETA, taxi to a CBP station and wait there until a CBP officer arrives.

On the other side of it, it looks like CANPASS would let you land in Canada even if the CBSA station is closed...

Edit: actually, one interesting thing to consider about routing vs. schematic design: my general experience with LLMs is that things work best when there's a feedback loop. On the schematic side, there isn't really much of a feedback loop; nothing will tell you that you've e.g. flipped the + and - terminals of your Opamp.

On the layout side, you've got the ratsnest and DRC to tell you whether you're winning or not. That might be the key to making LLM-based routing work. Opus refused to hand-route my whole board yesterday but is currently using the KCRT tools to do the routing and things seem to be progressing well.

Claude (web) saved me a ton of money with a new hobby. I was curious about astrophotography and had a budget to fill in any missing pieces from my existing photography kit. The conclusion was basically: "get a $20 intervalometer and a wall-wart battery[1]. The lenses you've got are fine, the tripod you've got is fine, and hey good news, Jupyter and Venus are near conjunction so if you can find somewhere dark where you can shoot to the west, you can get a pretty cool timelapse this weekend"

Later on also provided some rough guidelines for the partial lunar eclipse that resulted in some pretty amazing photos.

Since then I have been doing a bit of research on getting a real mount, so that initial budget will probably get used, but it'll be a) for kit that I genuinely don't have yet and b) spent after I've already confirmed that I thoroughly enjoy going out and photographing the stars for hours and then stitching all those photos into timelapses or other neat stacks.

[1] I'm not sure what the right name is for this. One end is battery-shaped but has a wire that plugs into a wall-wart instead of having to switch batteries. I had an existing 500-Wh 120V battery bank thing and it was more than enough to run the camera for 6 hours.

My first attempt on a low-speed not-dense board came out OK. There were a few odd choices, but I don't expect perfection from auto-routers. If nothing else, auto-routers are a good way to double-check that your floor planning is OK, even if you then rip up the auto-routed traces and hand route them instead.

Heh, tongue firmly in cheek: there's a good reason to change from I2C to SPI. I2C's pretty shit in comparison :D

More seriously, I've had interesting mixed results. I had a hobby-maybe-commercialize-next-year Christmas lighting project that I started in March (if I start in March, this will definitely be ready for Christmas!) and got 90% of the way there... and then life got busy and it's sat there untouched with the holiday season coming up again quicker than I'd like. I threw Fable at it first to assess what's done, not done, look for schematic mistakes/improvements (it rightfully pointed out that the cheap LDO I was using was near its limits and suggested a better part).

Opus did a reasonable job of swapping out the LDO, putting reasonable values on all of the pull-ups/pull-downs, etc. Some of the labels did end up in slightly weird places and needed some manual correction. It also did a decent job of floor planning on the PCB side, but pretty much refused to do any routing other than power traces.

I've installed https://github.com/drandyhaas/KiCadRoutingTools this morning after seeing this article last night... we'll see how this goes. My first manual attempt at running the router did decent but it failed a bunch of DRC spacing rules and insisted on via-in-pad to make it all work. That's, I think, a side effect of Opus's floor planning. It's interesting to play around with to see what's good and not.

Comes out to about $100/month

$3.2B revenue, 2.6M player accounts, ~$1230/yr ~= $100/month

So can I still drink and gamble in private if I sign a waiver?