Paul Irwin

Today's post Unit Testing Umami.Net - Testing UmamiClient This covers the start of the test suite I'm building for Umami.Net with the testing of the HttpClient based UmamiClient https://lnkd.in/eJvtNymv #xunit #umami

1

I've been trying to test Hardware stuff these days one thing I got my hands on is the Maixcam from @sipeed its a Risc-v based device. It's a camera that also does AI inferencing and people have been using it for various usecases such as tracking changes in plant growth etc. I have yet to try stuff out but here are the first few videos. more in the caption #riscv #ai RISC-V International

10

1 Comment

My thought today is on Global Collaboration. I moved countries 10 times between 1999 and 2019 (four times to the USA, once to France, once to Austria and four times back home to Scotland) I’ve worked in: France, Belgium, Italy, Netherlands, Germany, UK, Poland, Romania, Hungary and Ukraine. I’ve also worked in the USA, Mexico, China, Singapore, Malaysia and Thailand. WHY is Global Collaboration hard? The reality is there are a number of challenges: 1. Geography and Time Zone 2. Knowledge (functional,explicit and tacit) 3. Social & Cultural Anyone who wants to get something done in a global company will be faced with these challenges. How can you possibly succeed? I don’t have all the answers, but I’m happy to share some of my experience on what I have found to be helpful over the last 25 years and how you can help set your teams up for success. Stay tuned for my next post.

152

6 Comments

Chosen-prefix Collisions for MD5 and Applications Present a novel, automated way to find differential paths for #MD5. As an application we have shown how, at an approximate expected cost of 239 calls to the MD5 compression function, for any two chosen message prefixes P and P 0 , suffixes S and S 0 can be constructed such that the concatenated values #PkS and P 0 kS 0 collide under MD5. The practical attack potential of this construction of #chosen- #prefix collisions is of greater concern than the MD5-collisions that were published before. This is illustrated by a pair of MD5-based X.509 certificates one of which was signed by a commercial #Certification #Authority (#CA) as a legitimate website certificate, while the other one is a certificate for a rogue CA that is entirely under our control (cf. http:// https://lnkd.in/dSfdTP4d). Other examples, such as MD5- colliding executables, are presented as well. Centro de Investigación de Ciberseguridad IoT - IIoT

17

Nice to see the UK Government supporting the move to quantum computing with the opening of the new National Quantum Computing Centre. #NQCC #QuantumComputing #Encryption New national quantum laboratory to open up access to quantum computing, unleashing a revolution in AI, energy, healthcare and more

2

An interesting experiment with NotebookLM - I've been using it recently to extract useful points from papers, case studies, webinars etc - all the sources therefor are pretty well formed and coherent, but I wanted to see how good it is at extracting usable content from reaaallly bad inputs. I prompted ChatGPT & Claude to generate some really badly written texts about "the history of ice cream" - 3000 words each and to mix up the timelines, go off on long unrelated tangents and generally write like a drunkard. Both pieces are virtually unintelligible and hard to tell what the subject actually is, and I copy pasted parts of these into each other to make even more of a mess. I asked Notebook to generate the audio overview and it pieced together a completely coherent timeline - the only place it went wrong was understanding that 19th century is not the same as 1900's - but otherwise it produced a really accurate, decent and enjoyable summary of the hard facts with just a smattering of the tangential nonsense. Podcaster 1#: "The source is all over the place to be honest ....let's see if we can piece together this story" For me this is the most impactful / genuinely useful tool since ChatpGPT...

11

6 Comments

I look forward to using these new VisionFive V2 RISC-V boards in my Computer Architecture course. They look to be improved over the V1 boards

14

Heading to ISCT 2024? Our agenda is jam-packed with opportunities to explore and move forward with your advanced therapeutics — from exciting speakers to thought-provoking sessions, and educational posters you won't want to miss: 👉 End-to-end manufacturing of autologous CAR T cell therapies with a new cell therapy manufacturing platform 👉 Regulatory T cell expansion in the Xuri cell expansion system W25 👉 Closed, semi-automated harvest of iPSC scale-up workflow using the Sepax C-Pro with CultureWash protocol 👉 Optimization of multiplex CRISPR-Cas9 editing of human primary T cells using lipid nanoparticles (LNPs) and subsequent off-target evaluation 👉 Strategies for producing clinical and commercial RNA-LNP drug products And there's more...follow this link to discover all the ways you can connect with us at ISCT: https://lnkd.in/d9RUnti8

23

Exciting times: the Raspberry Pi Pico 2 is now official, featuring the shiny new RP2350 microprocessor - a part which boasts two Arm Cortex-M33 cores *and* two RISC-V cores based on the free and open-source Hazard3 core. Even better: it's backwards-compatible with the original Pico and RP2040, so existing projects - including the ones in my MicroPython book - should work just fine! https://lnkd.in/extydbT7

7

2 Comments

I've just ploughed through a 132 page report on AI safety so you don't have to (you're welcome!) and it offers a fascinating snapshot of how quickly the conversation has moved on. When I attended the world's first AI Safety Summit in Bletchley Park, in November, I asked whether short-term threats like job losses and bias were in scope for the inaugural discussions between world leaders and tech giants. I was told no, this would be an event focusing on the absolute worst case scenarios: the big, blockbusting, doomsday-style existential threats. Fast forward to May 2024 and the word "safety" has been removed entirely from the title of the gathering, which is now called the AI Seoul Summit (it is being held in S Korea). The latest report, which has just been published by the AI Safety Institute that was set up following Bletchley, says from the get-go that there is an "uncertain" likelihood of any doomsday scenarios. There is "no evidence yet" that gen AI could automate a sophisticated cyber attack or develop a lethal bio weapon, it goes on, and the risk of the complete loss-of-control of an AI system by humans has a "highly contentious" plausibility. I'm slowly climbing back out of my imaginary bunker. That said, the threats the Institute does focus on do rather resonate more with what I was saying six months ago. Issues with bias (the report asks "can AI ever be completely fair?"), copyright and the resulting lack of quality data to train systems on, job losses, and sustainability (AI is a power-hungry beast. The US "may struggle" with the AI sector's demand for electricity, it says). The report, written by AI "godfather" Yoshia Bengio who himself has warned in the past of existential threats, concedes that nobody, including AI developers, really understands why AI generates the output it does. There is currently no good enough way of doing it. It also states that the current go-to process of safety-testing AI tools by "red-teaming" them - that means assembling people to try to force them to do bad things, has no official best practice guidance. Red Team evaluators can display their own bias, it admits. Overall, bias and representation is described as "an unsolved problem" - unfortunately I think we all know that by now. The conclusion of this report isn't exactly a headline writer like mine's dream. The future is uncertain and nothing is inevitable, it basically says. Colour me not surprised! It will be very interesting to see how the conversation is shaped six months from now. Maybe I'll be back in the bunker by then....

800

77 Comments

Nearly scrolled past this, but it really got me thinking. What really sets a team apart? It’s not just about technical skills—it's about how well a team can communicate, especially when remote. And I don’t just mean work-related chat, but being able to talk about anything. Communication is a skill, and like any skill, it needs practice. It’s the foundation that allows a team to grow, collaborate, and learn. This ties in nicely with the whole remote vs office culture debate. For a team, there shouldn’t be much difference when it comes to staying connected. Tools like Slack huddles are a great way to keep those conversations going, and in theory, they’re no different than walking across the office to have a quick chat, you really have to maintain that ad-hoc culture within the team. #remote #culture #highPerformanceTeams

10

1 Comment

I really like sound of this version of 'Blue Monday', https://lnkd.in/dv-Y6ckX but there are many, many things about the video that annoy me. A small sample : * that's a cruddy monochord (I made one as a pressie for oor Nigel, much more practical & prettier) * the sig gen looks at least late 40s * voice short delay effect - how? It's generally taken that the first time an artificial reverb effect was used was in 1947. The producer's bathroom, for this : https://lnkd.in/dfYXnTwd  * where is the vocal chorus coming from? (Ok, magnetic wire recording had been around a while, tape was just emerging, you could overlay - but they were all really big machines, valve/tube electronics, I don't see one) * NO-ONE IN THEIR RIGHT MIND USES RUSTY STRINGS However, all these are offset for me by intrigue on the 4-stringed hammered thing (what is that called? Want!) and most of all by imagining - cue Stephen Morris drum roll: thud thud thud tss tss-tss... Peter Hook playing a banjolele. (I've actually got one, they are probably the most ridiculous instrument in human history. A meeting of the finest engineering and craftwork. The end result having the worst features of both banjos and ukuleles. I'm guessing me dad found mine in a charity shop. Has been desperately in need of repair for decades (both dad & banjolele), this might be motivation for me to get around to it. (Plus a mandolin I got from a car boot sale years back which shares the same cardboard box. That had no back and a broken front when I bought it for pennies, and which I spent many hours resurrecting. Done so badly it's needed rebuilding, gluing back together, since a week after my fix. Accursed kerfs!). Anyway, How does it feel? https://lnkd.in/dwsHJPSK How does it fee-el... https://lnkd.in/d9j72XvW

1

This blog post wraps up my short series on choosing FFT transform lengths when using FFTs to compute full convolutions: https://lnkd.in/ewDfq_w9 See also my MATLAB function fftTransformLength.m on the File Exchange: https://lnkd.in/eYqHBdnH Thanks to Christopher Turnes and Cris Luengo for their helpful input.

31

1 Comment

OnDemand moves super fast showing the future of the work environments. Making them more immersive and efficient.

1

Alastair captures one of the biggest problems with the current RISC-V spec. When I looked into this (formal spec and compliance) a while back for personal interest I was (almost) shocked at the lack of rigor and state of the tooling. I say almost because I expected it, but was still sad. What saddens me more is that this situation extends to implementation validation. The status quo is sufficiently non-rigorous that if you were building an implementation you would have to do most of the architecture validation tooling lifting yourself. This will result in a few very rigorous designs (from folks who get the problem), and at least some that are not (folks who make a mess). I expect as a result an escape in a large scale design that hits the market and breaks compatibility. This stuff really matters. Read Alastair’s blog here, then remind yourself there’s a reason why some specs are thousands of pages long:

50

4 Comments