I think there's a lot to be said...
For just learning to use and configure Vim or Emacs really well. Certainly Vim if you spend a lot of time talking to strange computers over a network.
1098 publicly visible posts • joined 24 Jun 2009
This is impressive. What a heart does is pretty straightforward in engineering terms, so any breakthrough in materials science/tribology could lead to something more practical and longer-lasting than a human heart transplant, which comes with all manner of risks and complications.
The most complicated things I've asked AI to write for me are bits of code to illustrate how concurrency is done or simple REST clients and servers in a new (to me) language. In this respect, I've found the LLMs I've used to be a kind of useful teacher that helps me get to a point faster than with traditional methods. With rare exceptions, I write the code that solves the problem! The exceptions? I occasionally need to automate something in MS Office and have zero interest in learning a dying macro language. And let's face it, some work (e.g. writing a resignation letter) is grunt work and might as well be done via AI.
And now the BT contract is up, I'm making preparations to leave (just waiting for a Cisco VoIP box to arrive so I can test with an A&A account). I will be paying significantly under half what I pay BT for Broadband + Landline + Mobile (and that is taking into account a price hike at the end of my first year with the Alt-Net).
I agree that what Nvidia makes will still be in demand. Last year, the talk was of LLM development hitting a plateau of diminishing returns, or a wall. The compute cost of o1 seemed to bear that out... DeepSeek potentially moves that wall into the distance. And it's FOSS. Anyone can take it, study its guts and improve it. I think it also points to OpenAI in particular being a massive fake it 'til you make it scam.
The compute cost of doing anything with OpenAI's most advanced model is horrendous, yet others get close with far fewer resources. As anyone who squeezed a flight simulator into 1k will tell you, scarce resources can drive efficiency and innovation.
Alongside Open AI. There’s more interesting and useful stuff floating around the world of AI (none of which will put developers out of the job*). Just ignore the shouty ones who are after unlimited money and resources.
* Well, accept at Meta where Zuck has been drinking the cool aid, I look forward to that.
"We always notice that it's the large companies that put out these mandates - it seems rare that smaller companies are discussed, except when they're citing larger companies."
Correct. My smaller employer stopped paying for an office during lockdown. I'm sure we'll have something small and reasonably priced at some point, but there's no stomach for paying rent on something large enough to take the whole team at any given time.
"Copilot+ PC is Microsoft's branding and specifies systems with an NPU that performs at 40 TOPS or more, where TOPS is one measurement of AI performance."
But what are they all for? If you use Chat GPT, Claude, Copilot or Gemini, online, those TOPS don't make a difference. Perhaps some of the cloud processing will be distributed out to local machines, but when? Far from being a killer application, Recall is wanted by nobody.
If I needed a new Windows laptop, tomorrow, I would probably buy a Copilot laptop. I don't need a new Windows laptop tomorrow.
Just get the user to identify the squares with effing traffic lights twenty effing times as I found when I wanted to change the password to my rather important* AOL account. Can modern AI really not prove it is not a robot with the current tests?
* My iTunes and later full on Apple ID from the turn of the effing century onwards. One of life's regrets.
Google has released Gemini with Deep Research which is similar to Perplexity but not as fast because it does more (and takes quite a bit longer).
I hope Perplexity finds a morally good way of becoming the new king of search. It excels at quickly coming up with a number of decent sources you can throw at Notebook LM type products. A Pro account comes with my currency card and I now use it more than Google for day to day searching.
Yes, the UK government were pretty stringent and required a letter from my employer. My reasons were:
- Lots of business in the Middle East (and possibly Israel also).
- Occasional need for a manager* to apply for a Visa on my behalf when I was already in another country.
- Later, business in Uganda and DRC (who don't get on).
Back in the day, it was nice to have the side benefit of turning up somewhere like the US with a passport with zero dodgy stamps.
At the next renewal, I will surrender my second passport (I don't need it anymore).
* I trusted one manager to keep the concurrent passport in his safe.
There will likely be a huge difference between businesses that use AI well and those infested with pointy-haired types who use AI without the vaguest hint of self-awareness. Perhaps we will see some sued to oblivion when they use AI-generated output that happens to breach A.N. Other's copyright.
Not quite as bad as that - just the directory I was in and all subdirectories. Bad enough that it was all of our software on a remote production system in Brentwood (over a 56k modem connection) - the customer in question did notice. The rather more experienced boss got me out of the shit that day.
Not by choice, but my employer wanted an Excel-based reporting tool that would query a back-end Informix database. The JavaScript-based alternative/replacement seems somewhat nicer, with none of the security blockers that make VBA impossible to run in some situations. I never thought I'd write that about JavaScript*. Modern Excel isn't too awful, and if, like me, you're not a power user, AI/LLMs can make surprisingly good coaches when wandering into the unfamiliar.
* I've never read a bit of JavaScript that didn't look hideous to me, but OfficeScript and its Google equivalent might convert me.
I wrote a hints and tips sheet on setting up WSL a while back and thought I'd have a play. I fed this to Notebook LLM and told it to produce a presentation. Now, it did have a limitation - it would produce the text of the presentation but not then chuck that into Google Slides. We're not quite there yet*. But I could also tell it to augment the simple instructions with benefits, drawbacks and use cases. Not bad - I have an extra training presentation for recruits to go with minimal effort.
Contrast this with the corporate Microsoft account... A lot of what Google Bundles costs nothing extra. Caveat: do not recommend feeding anything confidential or proprietary into any of Google's AI tools.
* Likewise, Gemini in Google Slides will produce a slide but not the whole presentation.
For test automation, it's worth trying Docker. I now have a Docker file that, when built, installs all the language and build tools, downloads the auto-test software from our repo, copies in the configuration for the system under test, and runs the suite of tests. I have another file for spinning up a dev environment.
That said, KVM, you are a step closer to OpenStack, and that's no bad thing bearing in mind VMWare's long suicide.
Having skipped a couple of generations as a very happy (none-pro) 6 user, the 9 crosses the line for me with its improved camera features, the main one being that 50 megapixels can now mean exactly that. The 6 can already just about replace a real camera if I must travel light, while the 9 Pro makes travelling lighter more appealing.
The AI features? I'm all for the ones I don't see (I'm all about the photography, and if an NPU makes that better, I'm fine with that), but the flashy, visible stuff leaves me nonplussed (though I'll doubtless find uses for some of it).