Juju-flop and turlingdrome I can countenance, but I draw the line at Belgium
Posts by Michael H.F. Wilkinson
4110 posts • joined 24 Apr 2007
Page:
Debugging source is even harder when you can't stop laughing at it
BOFH: Putting the gross in gross insubordination
BOFH: Gaming rig for your home office? Yeah right
Simon is on form, these days!
"... and even signed you up to host your mother's bridge club and offered to open up your garden to influencers' photo shoots."
Now that is pure evil, right there.
It's amazing what computers might do all on their own. Carl must feel so much safer with the BOFH watching over him.
BOFH: All hail the job cuts consultant
Beware the techie who takes things literally
Make assistive driving safe: Eliminate pedestrians
Stay in lane features can be a menace, I have found. Overtaking a cyclist typically does not require going to the opposite lane, but I do like driving around them rather than through them (personal preference, others may feel differently), as the lane assist would have me do. It is totally unnerving when the car (a rental Volkswagen Passat) basically tries to take over control of the wheel and nearly drives you into a bike. Don't get me started on the various alarms going off when trying to back out of a parking space, because the proximity sensor detects the presence of a wall in front of the car. A wall you are clearly moving AWAY from.
I am so glad my own car has far less AI (alleged intelligence) built in
BOFH: The Geek's Countergambit – outwitted at an electronics store
Trio of Rust Core Team members take their leave
BOFH: On Wednesdays, we wear gloves
Something 4,000 light years away emitted strange radio bursts. This is where we talk to scientists for actual info
BOFH: What a beautiful classic car. Shame if anything were to happen to it
Billionaires see wealth double during pandemic as tech bros lead the charge
Snap continues to make a spectacle of itself as it tries to trademark the word spectacles
On Christmas night, a computer logs a call to say his user has stopped working…
BOFH: The vengeance bus is coming, and everybody's jumping. An Xmas bonus hits me…
The Filth Filter is part of the chipset, honest. Goes between the TPM and SEP. No, really
BOFH: Time to put the Pretty Dumb F in PDF reader
BOFH: What if International Bad Actors designed the vaccine to make us watch more Steven Seagal movies?
BOFH: You drive me crazy... and I can't help myself
BOFH: So you want to have your computer switched out for something faster? It's time to learn from the master
I would drive 100 miles and I would drive 100 more just to be the man that drove 200 miles to... hit the enter key
Re: "So my urgent 200-mile round trip ended, …
Reminds me of the "IMPOSSIBLE ERROR" message I got from the wonderfully named D02BAE routine from the NAG Fortran Library Mark 14 on the Cray J932 way back when. I was running multiple instances of this ODE solver in parallel, and it turned out library had not been compiled for parallel execution properly (brilliant, on a 32 CPU shared-memory parallel machine). This meant only one instance of the the named common block used by the routine was created, so the 10 instances I was using were happily overwriting each others data, causing such confusion that an IMPOSSIBLE ERROR was generated.
Not a hardware fault, but it took some hunting down
AI caramba, those neural networks are power-hungry: Counting the environmental cost of artificial intelligence
They don't even do statistical analysis. If they did, they would be able to put confidence limits on their classifications at least.
The power issue is a serious one. Last year we published a "hand-crafted" method for stereo vision for a gardening robot that was perhaps not quite as accurate as some deep learning methods, but ran happily on a Raspberry Pi, as opposed to a powerful laptop with GPU. The accuracy was good enough for the purpose of the robot not running into trees.
Volkswagen to stop making its best-selling product for Wolfsburg workers: VW-branded sausages
BOFH: 'What's an NFT?' the Boss asks. In this case, 'not financially thoughtful'
Come fly with me. But first we need to find a boot device
Sentient Urinals?
I stood and wondered why the urinal in the motorway service station in Germany needed to reboot Android.
More cryptically, another such urinal had the message "Slave cannot connect to master ... Retrying"
Next up, urinals with the GPP feature, or defocused temporal perception, causing them to start sulking in basements
Without a trace: Baroness Dido Harding to step down as chair of NHS Improvement
BOFH: They say you either love it or you hate it. We can confirm you're going to hate it
BOFH: But soft! What light through yonder filing cabinet breaks?
Here boy! Making the Sample Fetch Rover that'll collect soil from the Red Planet
BOFH: Where there is darkness, let there be a light
BOFH: Here in my car I feel safest of all. I can listen to you ... It keeps me stable for days
DARPA nails cash to project 'FENCE' — a smart camera that only sends pics when pixels change
Potential problems: trees, leaves, wind and sunshine (or clouds)
This bears a striking similarity to a traffic camera system produced by a local IT company I visited a decade or so ago at least. The aim was to let traffic cameras only record the passing cars, and stay shtum when there was no traffic. So a simple image differencing method was implemented, and if the sum of absolute differences between consecutive frames was sufficient, the system started sending a burst of frames, until the situation became static again.
Some traffic cameras, happily transmitted continuously from sunrise to sunset, especially on windy, sunny days, when the pattern of the shadow and light caused by the sun shining through the trees along the road caused loads of pixels to change, without any actual vehicle or person passing by the camera. Rushing clouds could trigger similar problems, as could snowflakes, hail, or rain. In the end they had to do far more advanced object recognition, in particular recognizing license plates to make the system robust (pedestrians and cyclists were not of interest in this system).
No doubt the boffins and DARPA will have thought of this
Radioactive hybrid terror pigs have made themselves a home in Fukushima's exclusion zone
Study finds crayfish treated with antidepressants become more outgoing, adventurous
Hubble’s cosmic science is mind-blowing, but its soul celebrates something surprising about us
BOFH: Oh for Pete’s sake. Don’t make a spectacle of yourself
BOFH: When the Sun rises in the West and sets in the East, only then will the UPS cease to supply uninterrupted voltage
Poor cyclist
"The year Richard was born I set a company record for how fast a VAX 11/780 would roll down the road from our building. I could've got a world record if it hadn't been for that cyclist."
I did my MSc thesis work on a VAX 11/780. Big hulking beast, and everyone complained their vi or emacs editors slowed to a crawl whenever I was running thermal simulation software on it. I actually found my shiny new 80386 at 25 MHz with a Cyrix 387 co-processor and a whole 8 MB of RAM outpaced the big hulking beast by quite a margin.
Lovely episode. It was inevitable Dick was going to go for the practical exam
BOFH: Despite the extremely hazardous staircase, our IT insurance agreement is at an all-time low. Can't think why
BOFH: I'm so pleased to be on the call, Boss. No, of course this isn't a recording
BOFH: But we think the UK tax authorities would be VERY interested in how we used COVID support packages
Yep, the 'Who owns Linux?' case is back from the dead
Turns out humans are leading AI systems astray because we can't agree on labeling
I don't like saying i told you so, but ...
I frequently rail against claims made in AI papers that if the ground truth contains a percentage of errors, any AI system trained on them is likely to end up with a similar actual error rate. I have seen people claim an increase in performance from 97.6% to 98.1% (error bars not included) on data sets where there are two ground truths, drawn up by to medics, which are at odds with each other. In our own earlier work, we managed to get a sort of pareto optimum of 92.5 ± 0.6% on both ground truths, but were in places penalised for finding blood vessels the doctors had missed. It turns out, somehow ground truth 1 has been elevated to The Ground Truth, and the other demoted to "a human observer". And now AIs are better than the poor "human observer" simply because they have been taught to copy all the mistakes the other human has made.
If ImageNet contains up to 6% error, I will continue to take all claims of 99% or better performance with a considerable pinch of salt. Furthermore, if error bars ar not included, how can they claim to be better than an earlier method if the differences are sub 1%.
I am not saying deep learning and CNNs are useless, it is just that sloppy science does them a disservice.