Posts by tatatata 83 publicly visible posts • joined 18 Jan 2019
The question whether it is AI or real is a bit more difficult on LinkedIn. Some points to consider.
Recruiters and HR require a specific language and a specific set of incantations. Divert from that and you might not get invited for your next job. Do not write "use", but "utilize" , things like that.
"Thought leaders" are in general people with little or no actual skills. They subsists by copying each others "inspirational" quotes. It is therefore easy to confuse them with AI.
I once had, as a work phone, a recycled number from some for of internal legal desk of the police. People calling me in the night blurting out specific details about police actions before I could react. It takes a few of those calls before you understand what is going on. Called the police, changed number.
There are two good examples for using AI that I can think of.
Generating pictures to lighten-up your presentation. Not for content, just some illustrations.
Generating elevator music. When you use real songs in shops or pubs, you need to pay quite a hefty fee to play music. When you use AI generated mind numbing background music, no licenses are required.
Let's start coding:
While (some condition) {
<single keypress tab>do something
}
or
While (some condition) {
<pres a space><pres a space><pres a space><pres a space>do something
}
or even worse:
While (some condition) {
<pres a space><pres a space><pres a space><pres a space><press one space too many><backspace>do something
}
So for me, it is tabs.
Once again, an Agile apologist reacts to an observation about the bad side of agile. And guess what: it is not Agile's fault. Here the observation is how Agile kills innovation. Of course it does. Every engineering methodology kills, or at least maims innovation.Waterfall did, extreme programming did, and Agile too. Innovation IS locked in the frozen meat store of Agile. No engineering method is a substitute for thinking.
Ofcourse, there is the car-analogy, because everybody accepts that programming is the same as making or driving cars. "Devops can be a car crash, but car crashes are rarely the fault of the car."
The problem is "None of that will help if you’re still trying to fix old problems in the old ways. That’s been done." And Agile, dating from 2001 has become an old way of doing things.
And yes, if people understand and can work on a full stack, like security engineers that have been exempt from scrum and agile, that would help. It would make large projects harder though.
Nor do they seem to understand what they're doing.
If I was on a phone following a script, and the script told me to inflict some NSFW language on a potential customer, I would not follow that part of the script.
Unless, of course, those callcenter employees could not care less either.
The main issue here is bandwidth. If you want a picture of all the Windows users' desktop every second, Microsoft will probably need a lot of incoming bandwidth. And even if they don't want to use it, there will always be some who decide to put in on the Onedrive. I think the famous abbreviations in the US would like a copy too, because of, eh, terrorism. No? Childporn then.
No. it is not just you. It is one of the reasons I dislike Python. I have, after 40+ years, finally got used to the space/tabs in the Makefiles. I am also not a fan of YAML.
And for those who seem to be ok with these whitespace-horrors, please try to write anything usefull in Whitespace (the programming language). Or anything at all.
The benefit is clear.
Cloud costs are variable costs, without fixed assets. On-prem is fixed costs. For shareholders, it is very important that the fixed costs are as low as possible.
Nothing to do with IT; the only benefit is in bookkeeping.
C-suite (including RedHat's as I read recently) will turn to mcKinsey, who states that it is fine to "unleash developer productivity with generative AI", especially in four key areas:
- Expediting manual and repetitive work
- Jump-starting the first draft of new code
- Accelerating updates to existing code
- Increasing developers’ ability to tackle new challenges
Those who code should document really good where management pushed them into using AI without all the additional checks. It may become an issue in court.
Ah, but then, we expect IA to provide us with fault-free software that does what we want it to do.
When was the last time some company you hired, or your internal team of software developers produced fault-free software that does what you want it to do? AI can make buggy software much faster!
The problem is not whether or not the bulbs in the light fittings are replaceable. The fact that the unit can be replaced makes it possible for a car to be repaired. You don't have to throw away the car when a headlight breaks.
Of course, you might want to refurbish parts that you use to repair the car. That becomes even more important when a car gets older. At some point in time, those nicely shaped lights from your brand new car will be out of stock, and then, if you cannot refurbish parts, you may need to get rid of your car when the headlight breaks.
And that is just the windows PC. Many Android phones become unusable when the banking app requires a minimum version of Android, but the vendor decides not to provide that for older phones. Same for Sat Navs: older type of devices don't get map updates anymore.
It is not so much the right or obligation to repair; it should be the obligation to support: Every vendor of a device should support 15 years of updates of software to keep the device operational and functional after first sale to a consumer. That would limit the landfill.
As for Windows: I would love to see the discussions between Dell and Microsoft about 15 years support after last sell.
"There are cases where a very large LLM might require 32GB of RAM," he noted. “Everything else will fit comfortably in a 16GB system.”
Seems a bit like:
I have to say that in 1981, making those decisions, I felt like I was providing enough freedom for 10 years. That is, a move from 64k to 640k felt like something that would last a great deal of time. (Gates)
32 bits, it is enough for an experiment, it is 4.3 billion terminations -- even the defense department doesn't need 4.3 billion of anything (Cerf)
"I think there is a world market for maybe five computers." (Watson)
(and so on)
I think no one with any knowledge of AI is surprised.
These researchers generally pic something obviously wrong about AI, feed cases into AI and then prove something everybody knows. It is not research, it is publishing a paper to get attention and perhaps more funding. You might argue, that it could be a test-case before deployment, but it is not research.
I could propose another "research". The conclusion would be that AI is capable of answering simple coding questions, but fails when more complex programming questions are asked.
Up to 2020, everyone worked in the office. That was normal. So, working in the office is not some wild idea that Dell suddenly proposed. People did it before and will do it after COVID.
It is also not new that Dell wants workers in the office. Stopping WFH-allowances last year was also not the first indication. Dell is and has always been keen on getting people in the office. I think it is a bit naive to think that that would suddenly change.
That being said, it is quite ironic that Dell makes quite a lot of money from people working from home. Their website offers "Remote Worker Solutions", "Work from home services" and proposes to "Accelerate employees’ productivity and collaboration from home with Dell Work from Home Solutions." Next time you speak to a Dell-salesman, ask about these solutions, and ask if they are using them too.
The scary bit is not that CBDC can be programmed or earmarked to be used for a specific purpose. That will only hamper its adoption. The scary bit is, that even when the CBDC was not created with that programmability, it can easily be bolted-on later.
This is something to keep in mind when you read the EU draft regulation; "The CBDC “shall not be programmable,” the text added, following concerns that giving the ability to control how given funds are used could limit the freely usable nature of fiat currency. Basically, it means for the time being, the CBDC will not be programmable.
Unfortunately, techniques like LOCKS already exist. LOCKS (Locally Operated Cooperative Key Sharing) shares the session keys with a "trusted agent" via a modified NSS library. It is feasible that that "trusted agent" would be your favorite governmental body.
I can see different approaches how to subvert this system, from sending encrypted files to posing as trusted agent. But the fact that this is actively being developed worries me.
"The IPv4 address space uses 32-bit numbering .... "
It is sad that even on the Register it is deemed necessary to explain this difference between IPv4 and IPv6.
That being said, I recently switched ISP and now I no longer have direct IPv6 access to the Internet. The problem is, that many ISPs still do not provide IPv6. And why should a company that already has IPv4 addresses invest in a technology that only a small part of its customers can use and costs a lot to switch? Only if all ISPs offer IPv6 and many ISPs use CGNAT there is a slight chance that some companies might consider investing in IPv6.
“While uses cases have yet to be fully articulated, interest in the category is already strong. AI PCs promise organizations the ability to personalize the user experience at a deeper level all while being able to preserve data privacy and sovereignty.”
Although I'n not a native speaker, I always thought that my English was quite sufficient. But this to me is just a string of marketing terminology generated by a BS-generator. Other than "no-one knows what to do with it", I see no meaning of theses words.
The problem is, that most exams and certifications have become about memorizing specific things. That is not useful.
When I was young (grandpa speaking...) you learned the basic principles and had an exam about whether or not you were able to apply then. You learned about differential equitations, and on the exam, you proved that you were able to solve a problem. Of course, Chatgpt is able to solve the kind of sums that we had on the exams. That does not prove that the exam is broken. When we learned about computers, we learned about binary calculation, basics of information theory, the basic idea of networking, differences between SNA and TCP/IP and that kind of stuff. And yes, the exams we had could also be done by ChatGPT.
We learned the principles and how to apply them.
Nowadays, you do not learn the principles anymore. You learn lists or fixed procedures. And then, an exam is just testing if you learned the lists or are able to apply the fixed procedures. I've done CISSP, and that was the exam: learn a book from beginning to end and you'll pass. MSCE, CCNA, RHCSA: learn the book and you'll pass, even without understanding. Here, we call these exams "American style".
It's not the exams that are the problem, it is the education for the exams.
I marvel at the idea to "cripple AI chips". Does the US export only chips with an AIQ (artificial IQ) below 60? Do we force the chips to use artificial marijuana?
Are there any AI chips? no. There are NVIDIA chips that are used for AI. Sure, NVIDIA and INTEL can create GPU's with lower memory and bandwidth. And the US can stop export of the high performance GPUs.
It may seem nitpicking, especially on El Reg. But the problem is that Biden cum suis do have a view on AI that does not differ too much from the joking remarks I made above. And that is scary.
The Indian outsourcing companies were very popular because they were very cheap. Partly because of cheap labour in India, partly because of tax reasons her in the west (in the Netherlands, 30% of the income of foreigners used to be tax-exempt for 8 years). Good programmers from India are no longer as cheap as they used to be. Tax benefits are diminishing. So, the profit of the big companies are down.
Together with the observation that IT personnel that knows the environment can do a better and quicker job than remote personnel that just execute tickets, I'd say that things will get worse in India.
The story that the US Copyright Office tells feels a bit different from what I read in the article:
"In her application, Ms. Kashtanova listed the author of the Work as “Kristina Kashtanova” and stated that she had created a “[c]omic book.” The application did not disclose that she used artificial intelligence to create any part of the Work, nor did she disclaim any portion of the Work"
"Shortly after registering the Work, the Office became aware of statements on social media attributed to Ms. Kashtanova that she had created the comic book using Midjourney artificial intelligence. Because the application had not disclosed the use of artificial intelligence, the Office determined that the application was incorrect, or at a minimum, substantively incomplete. "
In other words: the application by Ms. Kashtanova was a bit misleading.
The discussion about the use of AI also refers back to the discussion of photography:
"[...] if photography was a “merely mechanical” process, “with no place for novelty, invention or originality” by the human photographer, then in such case a copyright is no protection."
So, if you want to copyright an AI-novel, you must show that there is some novelty, invention or originality. At the moment, AI is seen as a merely mechanical process.
Theft is the taking of another person's personal property with the intent of depriving that person of the use of their property.
No-one that infringes on your copyright intents you to deprive you of the use of that image. Copyright infringement is therefore not stealing.
It is not taking money away from you. You never had the money, so it cannot be taken away from you. If they had decided not to use the image, because it's licensed, you would not get the money anyway.
If you see someone driving through a red light, you would not call it "speeding". So please stop calling "copyright infringement" "stealing".
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
