@ovid@fosstodon.org
@ovid@fosstodon.org avatar

ovid

@ovid@fosstodon.org

Well-known software developer. American living in France.

I have a poetic license to kill.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

ovid, to random
@ovid@fosstodon.org avatar

Having trouble keeping you house warm during the winter? Consider fine-tuning an LLM.

I'm not used to my Apple Silicon Mac ever getting warm, or the fans coming on.

ovid, to random
@ovid@fosstodon.org avatar

I think one the best things about AI today is I'm hearing a lot less about blockchain and cryptocurrencies.

That alone is a vast improvement in my quality of life.

ovid, to random
@ovid@fosstodon.org avatar

I really should put together an introductory course on how to use ChatGPT effectively. There are some amazing techniques you can use to get great productivity out of it.

ovid, to random
@ovid@fosstodon.org avatar

I should change my name to #Perl Horber.

ovid, to ai
@ovid@fosstodon.org avatar

I've noticed that when I share related material ANYWHERE, I get less engagement. I sometimes get outright hostility. If I respond to AI-naysayers anywhere, I try to give detailed, thoughtful answers, always with personal experience and often with resources to back up my claims.

The hostility to generative AI is real and the outright denialism is amazing. People are regularly making bold claims that were true a year ago, but not today.

Why does opinion trump evidence?

ovid, to ChatGPT
@ovid@fosstodon.org avatar

At university, my speech professor marked me down several times for not summarizing my speeches if my summaries didn't begin with the words, "in summary" or "in conclusion." He insisted that without those words, the audience wouldn't recognize the summary! o_O

Today, reading an article with "in summary" or "in conclusion" almost always sets off my alarm.

ovid, to ai
@ovid@fosstodon.org avatar

Markov Chains are the rigidity and uselessness of Immanuel Kant's Categorical Imperative.

Transformers are the flexibility of John Stuart Mill's Utilitarianism.

I'm sure AI researchers and philosophers can both despise me now.

ovid, to random
@ovid@fosstodon.org avatar

Help me understand?

More than once I've gone through airport security and been told off by security officers for having liquids in a clear bag, but with translucent letters.

They then proceed to put that bag inside a completely clear bag and wave me through, as if all is now right in the world.
Putting "laziness," "idiocy," and "jobsworth" answers aside, is there some practical reason for them doing this that I don't know about?

ovid, to python
@ovid@fosstodon.org avatar

I really wish had the magic methods from . This would make so many problems in Perl much easier to solve.

ovid,
@ovid@fosstodon.org avatar

@enobacon I don't like the list of special names either. We could find a cleaner solution, but overload.pm is showing its age. There's been discussion about cleaning it up and offering a new version.

ovid, to ai
@ovid@fosstodon.org avatar

IT people: want to keep your job safe from #AI? Fight to keep our profession licensed. We've fought against that for years, but if you fight for it, you'll have a license that can't be legally automated.

It won't protect all IT jobs, but it will protect some.

However, I suspect that we won't bother. As jobs gradually slide away, we'll argue and dither and ultimately do nothing.

ovid, to random
@ovid@fosstodon.org avatar

One of the biggest issues with LLMs today is their massive power consumption needs, along with performance. To entrepreneurs, these are opportunities, not problems. As a result, there are two new developments in AI which you should know about.

The first is silicon-photonic (SiPh) chips. They use light instead of electrons to perform mathematical calculations. Details are scant, but they appear to be faster than GPUs & use less memory. https://blog.seas.upenn.edu/new-chip-opens-door-to-ai-computing-at-light-speed/

1/2

ovid,
@ovid@fosstodon.org avatar

It's https://groq.com/ which is more exciting for now. They've built the Groq Tensor Streaming Processor (TSP). GPUs are powerful, but they're not optimized for modern AI usage. The TSP is. Groq answers your queries at an astonishing 500 tokens per second. Instead of watching words pop up on your screen one at a time, like you were reading some 1985 BBS, the answer is almost instantly there. Check them out!

And they use less energy, too.

2/2

ovid,
@ovid@fosstodon.org avatar

@kellogh Ooh, can you share a link about the physical motion one?

(I'm pretty sure that someone out there is looking at biocomputing, perhaps with organoid intelligence, to solve this issue)

ovid, to OpenAI
@ovid@fosstodon.org avatar

I absolutely need to write about 's . It scares the hell out of people and this time it's quite rational. Sure, it will put people out of work. Technologies been doing that for as long as it's existed.

But who's going to be the first person put on trial for a crime whose main evidence is an generated video of them committing a crime?

The misinformation will be staggering, from political discourse, undetectable deepfake porn vids, and more.

timbray, (edited ) to random
@timbray@cosocial.ca avatar

Dear world, I need to hire someone expert in installing & configuring some open-source software on MacOS. Currently my blogging system breaks every time I type "brew upgrade" and I am out of patience. [EDIT: Problem solved.]

Tl;dr: I need to arrange that this 2-line Perl program works and survives updates:

use DBI;
use DBD::mysql;

Will pay an appropriate hourly rate. Private-message or email me.

ovid,
@ovid@fosstodon.org avatar

@timbray Typically, you don't need to use the DBD directly. If you connect to DBI, it will figure out which DBD you need and load it for you. In fact, this is so common, that I'm unclear why pre-loading it like that would cause an issue.

Without seeing the error message, it's hard to say more. I'm not looking for a contract, but I'm happy to offer advice.

ovid, to random
@ovid@fosstodon.org avatar

I just watched the Shardlake trailer, only available on Facebook, of all places.

If you're not familiar with the historical Matthew Shardlake novels by CJ Sansom, they're god-damned amazing. @disneyplus made them look like god-damned pathetic.

The series might be incredible, but it's been a long time since I was so excited to find out I could watch something, only to have a trailer tell me "no."

This is why DisneyPlus has such a shit reputation. I suspect cocaine was involved.

ovid, to llm
@ovid@fosstodon.org avatar

I may have stumbled on a valid use-case for hallucinations.

Experimenting with fine-tuning LLMs for chatbots for for a post-apocalyptic society where everything pre-apocalypse is a rumor.

One NPC mentioned they heard that Austin had been housing an underground resistance movement for an uprising against [the antagonist]. Absolutely not true, but it's ... perfect. We want this kind of misinformation.

ovid, to python
@ovid@fosstodon.org avatar

: Batteries included

: Flint and tinder included

: PHB included

: Notepad included

#C : Screw you

: Arcane eldritch data guardian forbidding memory blasphemy included.

ovid, to random
@ovid@fosstodon.org avatar

A mini-rant about IT project management and those who approve IT projects.

There is this pernicious belief that being a PM is about delivering a project on time and on budget. The people who believe this often also naïvely believe that being a manager is about giving orders. They're dead wrong on both.

The dirty secret: costs and deadlines are not the droids you are looking for. 1/11

ovid,
@ovid@fosstodon.org avatar

Wait, what? Of course project cost and schedule are important! If I have a 250K budget and the project cost is estimated at 1M, it will never get approved. If I have a legally mandated deadline, schedule is important. But these are the exceptions, not the rule. For the rest of this, assume that schedule == time and time == project costs.

So if we aren't looking at project costs, what should we be looking at to build value?

NOW you're asking the right question. 4/11

ovid,
@ovid@fosstodon.org avatar

You know how to judge the value of something? You get real, meaningful data. In the case, for example, of dealing with , don't listen to politicians. Their goal is to get reelected. Don't listen to Red Cap guy who's watched too many YouTube videos. Their goal is ... well ... I'll let you be the judge of that.

Listen to the insurance companies. Their goal is to turn a profit and they have to use real data. 5/11

ovid,
@ovid@fosstodon.org avatar

As it turns out, many insurance companies have decided to no longer issue homeowner policies that are likely to be impacted by climate change. These are not tree-hugging hippies. These are business people whose companies will fail if they don't use real data.

https://ehsdailyadvisor.blr.com/2023/06/insurance-companies-decline-coverage-due-to-climate-change-risks/

How do they do this? They employ actuaries. 6/11

ovid,
@ovid@fosstodon.org avatar

Actuaries are professionals who use hard data and lots of math to predict the probability of future events, handle enterprise risk management, calculate the reserves needed to meet future claims, and so on.

Enter Douglas Hubbard, a man with such deep experience in this field that two of his books are required reading for the actuaries exam of the United States. https://en.wikipedia.org/wiki/Douglas_W._Hubbard He's very, very good at this and he's been studying what makes projects successful for years. 7/11

ovid,
@ovid@fosstodon.org avatar

Douglas Hubbard has written and talked about the problem extensively, but he's found that most project estimates focus on costs, not benefits, and as a result, they get their value calculations wrong. He calls this "The IT Measurement Inversion." https://www.cio.com/article/274975/it-organization-the-it-measurement-inversion.html 8/11

ovid,
@ovid@fosstodon.org avatar

In reality, the two biggest concerns you have for a project are, in order, the likelihood that it will be cancelled and the rate of project utilization (this is the value!) When was the last time you included the chance of project cancellation in a project estimate? This is why investors want to see financial projections. They know the numbers are often found by a proctologist with a flashlight, but they're focused on potential value, not just potential risk. 9/11

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • anitta
  • InstantRegret
  • mdbf
  • ngwrru68w68
  • magazineikmin
  • thenastyranch
  • rosin
  • khanakhh
  • osvaldo12
  • Youngstown
  • slotface
  • Durango
  • kavyap
  • DreamBathrooms
  • JUstTest
  • tacticalgear
  • ethstaker
  • provamag3
  • cisconetworking
  • tester
  • GTA5RPClips
  • cubers
  • everett
  • modclub
  • megavids
  • normalnudes
  • Leos
  • lostlight
  • All magazines
    •