This article tackles the issue from a job opportunity perspective, but a bigger problem is the quality of students completing CS degrees is declining. UC Berkley are seeing it in their STEM departments [1] and I have seen the raw data for other Universities delivering CS degrees that is unpublished.
Currently the only method to stop students from cheating is to run strictly controlled paper-based exams, and with smart glasses with built in LLMs, this is becoming more and more problematic. Anything not run under strict conditions is entirely untrustworthy.
Management is slow to catch-up or react and the lecturers running these degree courses are under significant pressure to increase the results. I'm aware that many are doing class-wide weighted adjustments just to keep the numbers of passing students up. The quality of students graduating with CS degrees is declining rapidly.
> a bigger problem is the quality of students completing CS degrees is declining. UC Berkley are seeing it in their STEM departments
The decline in quality of STEM students/graduates is alarming, but the decline in intellectual quality of students is generalised.
Little did Dodson think he was being prophetic when he wrote satirically of Reeling, Writhing, and the arithmetical operations of Ambition, Distraction, Uglification, and Derision.
[0] _ Lewis Carroll, Alice's Adventures in Wonderland
Isn't this just grading on a curve, which has been done probably as long as universities have existed? The key is the instructor making sure a high standard is met (which seems to be the crux of the issue).
Yes it has been in practice for a long time, but it's now being used to push clear fail cases into passing grades just to meet quotas. Prior it was used to adjust for particularly difficult assessments, and was closely monitored.
Less competition for me, and "educators" are being punished HARD for their abrogation of their actual responsibilities, which was to teach and give exams.
All exams should be verbal. The fact that verbal exams are so rare is because teachers/professors are overworked and (outside of AI) underpaid. Too many students, not enough time.
The moment you pull up a powerpoint and start reading off of it, or start assigning homework, you've already failed to implement the traditional liberal arts education that the humanities seems to fawn over so much.
There's ACTUALLY no solution to blooms two sigma problem (https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem) except for teachers to fundamentally change their responsibilities. More time needs to be spent being intention to every individual student. If that means we need fewer students in universities, so be it. AI will kill the impenitence for higher education anyway.
> Less competition for me, and "educators" are being punished HARD for their abrogation of their actual responsibilities, which was to teach and give exams.
Universities consist of a wide range of people with different incentives, the lecturers typically (in my experience) have very pure motives. It's the management parts that put pressure to pass students, meet metrics, etc.
> The moment you pull up a powerpoint and start reading off of it, or start assigning homework, you've already failed to implement the traditional liberal arts education that the humanities seems to fawn over so much.
Homework is essentially dead post-LLMs. The lecturer's responsibility is to provide guided learning, but also most importantly to assess each student's attempt to learn.
> There's ACTUALLY no solution to blooms two sigma problem (https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem) except for teachers to fundamentally change their responsibilities. More time needs to be spent being intention to every individual student. If that means we need fewer students in universities, so be it. AI will kill the impenitence for higher education anyway.
You'd be surprised how much 1:1 with students there are. One example I'm aware of is CS students getting 4 hours 1:1 for one module per semester - that's a hell of a lot.
What you're ultimately up against is cost per student. The overheads in Universities are enormous. It's usually 40:60:+, so £40k pay, £60k overhead plus research and investment (conference paper, travel, journals, new tools, etc).
I think Google will win out in the end. They are concentrating on what matters, performance per watt, and performance per dollar. They are building their own inference hardware and are working towards edge-computing which removes latency and compute overheads. These big LLMs are not yet cost effective, Google is just letting them burn their investment funds to "sell" to consumers at below cost.
After the AI bubble bursts, it will be the likes of Google that come out the other side still wearing their shirts. I think this bubble is out to scalp some giants.
I'm surprised nothing more serious happened. There was obviously a serious electrocution risk, but I think that is the easiest bit to deal with. 100kW of radio waves, whilst non-ironising, can still microwave you. With 100kW there could have also been a serious reflection back to the transmitter. This guy cutting the cable is far luckier than he will ever know.
Putting aside there's a good chance that it was written by an LLM...
Whilst I don't personally see myself in this story, I think I see others around me in it. Over the weekend there was some calculations that needed to be done, and the people around me were impressed with how quickly they were done in Libre Calc. They had trusted an LLM to do the calculations for them, which turned out to be wrong - because of some details that an LLM couldn't have possibly have understood.
I see people "generating ideas" which are clearly flawed in ways that I have to explain multiple times. Approximately 40 people were tasked to generate ideas, and somehow they all magically came to the same ~10 ideas, which they claim were originally theirs.
The other day I was in a meeting to discuss one such idea, and the person was explaining it by reading the LLM output from their phone. When I pressed them for more details, they then had to ask the LLM to elaborate. They were completely unable or unwilling to engage in any kind of thought process.
I'm no longer sure if I'm somehow getting smarter, or people around me are getting dumber. The AI bubble crash couldn't come faster.
> Adyen will take over GOV.UK Pay card payments for local authorities, police forces and armed forces units from Stripe, as well as pay by bank services, under a three-year contract worth up to £25.3 million.
I would prefer they take this money and either build a payment processor or use an existing UK company. The UK government is addicted to offshoring all contracts it can, and then is surprised when the cheapest possible quote actually ends up ballooning over the agreed amount.
Risk, time, complexity, mismatched skill-sets... "getting a new thing built" didn't work so well with (for example) Fujitsu and the Post Office, or the billions spent to little or no avail on NHS digitalisation. Seems to be a case of "damned if you do, damned if you don't".
> use an existing UK company
Are there established UK-based payment processors with equivalent abilities?
> Risk, time, complexity, mismatched skill-sets... "getting a new thing built" didn't work so well with (for example) Fujitsu and the Post Office, or the billions spent to little or no avail on NHS digitalisation. Seems to be a case of "damned if you do, damned if you don't".
These projects are possible, it's just clear the current project management structure of the UK government is not up to the task. I would lean heavily towards a UK-based private sector solution.
> Are there established UK-based payment processors with equivalent abilities?
This is the point, there should be one, and the UK government could heavily put their thumb on the scale to ensure that payments are processed within the UK.
It seems mad to have foreign companies make money on UK government fund raising activities.
If you ever worked with any of the UK gov or agencies you’d understand why it happens. Despite having its own IT in every agency those same agencies buy externally because “it is too difficult”. IT is entangled beyond belief and some of the knowledge is institutionalised. Existing contracts do not make it easier either. Any high ranking clerk is afraid of any risk associated as well. Not to mention the procurement process is hard and lengthy.
I understand, but it's why things are desperate to change. I refuse to believe that the UK is fundamentally unable to rise to these challenges, in which can I can only chalk it up to poor government structure and management. These people have a lot to answer for.
Why is that so insane? Would you be saying the same when the NHS was originally proposed?
It's entirely possible, the fact that the UK government lacks imagination is at great cost to the tax payer. There is no serious long term investment into UK tech, past building AI datacentres for a bubble economy with some of the worst energy rates in the world.
Typically you have -1/N for incorrect selection, where N is the number of choices. For N=4, you would grade incorrect answers as -0.25.
If you have a person taking an exam that is not confident in themselves or generally knows the subject area, you don't want to negatively impact educated guessing.
> Typically you have -1/N for incorrect selection, where N is the number of choices. For N=4, you would grade incorrect answers as -0.25.
That is definitely not typical. -0.25 is the appropriate adjustment for N=5. For N=4 you want -0.33. -1/N makes no sense at all.
Note that doing this preserves the expected value of everyone's score, but artificially widens the variation, which you might not want. It does allow you to diagnose partial knowledge, which you probably do want.
Maybe it is typical only for me. I did question it and was told that we don't want to completely remove the incentive for educated guessing. We also usually have a scale of question difficulty, so getting people to a pass is not too difficult if they know the subject at all, but getting towards 100% gets significantly harder.
I think the real reason is that our questions are usually N=4, negative marks of 0.25 allows for quick adding.
> I did question it and was told that we don't want to completely remove the incentive for educated guessing.
So... you were told some unmotivated nonsense?
On an item with four answers, +1 for a correct answer and -0.25 for a wrong answer means that in expectation you will receive 0.0625 points for a completely uneducated guess. The only correct adjustment you can make is to dock 0.33(3...) for a wrong answer, in which case an uneducated guess is worth 0.0000 points and a minimally-educated guess, one in which you're capable of eliminating just one of the four answers, is worth... 0.0833(3...) points.
> I think the real reason is that our questions are usually N=4, negative marks of 0.25 allows for quick adding.
You think adding fourths is easy, but adding thirds is hard? If you really believe that, it'd be simple enough to add fifth choices to your questions.
Are you sure the real reason isn't just that nobody ever bothered to put any thought into what they were doing?
> It is concluded that Rust is a sound choice today for firmware development in this domain.
This conclusion was reached with a single experiment.
> Two teams concurrently developing the same functionality — one in C, one in Rust — are analyzed over a period of several months.
> Furthermore, Ariel OS is shown to provide an efficient and portable system runtime in Rust whose footprint is smaller than that of the state-of-the-art bare-metal C stack traditionally used in this context.
> The authors thank Davide Aliprandi and Davide Sergi of the STAIoTCraft team, and the wider Ariel OS team.
So one team had Ariel OS developer support, and it's unclear what support the other team had. Seems fair.
In Figure 12, they simply stop optimizing the code once desired rate is reached. Just at the end of the project the Rust firmware gets over a third performance boost, most likely from their OS developers.
Additionally, there is a claim that "Ariel OS is shown to provide an efficient and portable system runtime" - but there are no real tests for portability are conducted. Worst still:
> Where C-based projects require a separate project setup and manual code copying per target, Rust on Ariel OS consolidates everything within a single project [..]
This claim is just not true. This sounds like somebody that is not as familiar with C.
> In Figure 12, they simply stop optimizing the code once desired rate is reached.
Yes. The goal was to handle the maximum data rate of the used sensor, and stop there. Time was limited on both ends.
> Just at the end of the project the Rust firmware gets over a third performance boost, most likely from their OS developers.
The ST intern found those boosts all by himself. They compared the exact MCU & peripheral initialization of the C and Rust firmwares, tightened I2C timings (where STM Cube has vendor tuned & qualified values), and enabled the MCU's instruction cache, which somehow is not default in Embassy's HAL. We were quite impressed actually, the last days before the deadline were quite productive, optimization wise.
> Yes. The goal was to handle the maximum data rate of the used sensor, and stop there. Time was limited on both ends.
I understand, and I understand that there were limits to what could be done with the resources there were. What irks me is the strength of the claim made without enough evidence to make it.
> The ST intern found those boosts all by himself. They compared the exact MCU & peripheral initialization of the C and Rust firmwares, tightened I2C timings (where STM Cube has vendor tuned & qualified values), and enabled the MCU's instruction cache, which somehow is not default in Embassy's HAL. We were quite impressed actually, the last days before the deadline were quite productive, optimization wise.
Fair enough, hats off to the intern. This kind of thing is common in MCUs, even on low-end CPUs weird defaults can be selected. But the involvement and influence of the OS developers remains unclear.
Again, there's just not enough data to make such strong claims. I think the paper could easily make recommendations, it could say that at least in some cases (as evidenced) Rust could be a reasonable choice, and it could make an argument for further work.
> This conclusion was reached with a single experiment.
No shit. This is the conclusion reached at the conclusion of this experiment. This part of your comment can be removed with no loss of clarity, I think.
I think you miss my point. I don't think that this conclusion can be reached with the (singular) experiments performed because there is a lack of data to draw it.
If I ran an experiment where I gave a cancer patient bread, and then they recovered from cancer, I couldn't then say: "It is concluded that <bread> is a sound choice today for <cancer treatment> in this domain.". You would rightfully jump up and down and demand further experiments to increase the confidence of the result before drawing the conclusion.
It could have been concluded instead that there is a case for further experiments to be conducted, or that Rust could be approaching a maturity where it could be considered for some firmware projects. But as it stands, the conclusion is far too strong given the experiments performed.
Currently the only method to stop students from cheating is to run strictly controlled paper-based exams, and with smart glasses with built in LLMs, this is becoming more and more problematic. Anything not run under strict conditions is entirely untrustworthy.
Management is slow to catch-up or react and the lecturers running these degree courses are under significant pressure to increase the results. I'm aware that many are doing class-wide weighted adjustments just to keep the numbers of passing students up. The quality of students graduating with CS degrees is declining rapidly.
[1] https://www.dailycal.org/news/campus/academics/failing-grade...
reply