Comments by "Michael Nurse" (@michaelnurse9089) on "Continuous Delivery" channel.

  1. Computer science and software engineering should be different degrees. With software engineering you should learn about dependency injection, client specifications and continuous deployment all in the first week. In computer science you can learn about ENIAC, Scratch and calculus all day long, if that is what you want.
    23
  2. I have worked a little bit on both sides of this issue - management and developer. The thing with the concept of technical debt is it assumes a stable business environment which is very far from reality in most startups. You need to communicate with non-technical managers using metaphors about what you are building. The best metaphor is the building industry. Some people build building only designed to last a week - for a trade show for instance. You need to be clear that 'duct-tape software' needs to be rewritten and refactored before it is anywhere near stable and that this takes MUCH longer than writing the first version of it. If they say 'there is no budget for that' then you should probably resign the next day and go somewhere with better understanding of what programmers need to do.
    10
  3. I think you are missing the whole point. Overloading employees is a discussion separate from whether you plan your work 'top-down or bottom up' which is what the whole agile/kanban/waterfall discussion is all about. Of course, the get conflated because planning can be come a tool of corporate dysfunction. Hard scientific evidence shows that after a couple hours employees start to actively harm results by just being there. Errors, boredom, resentment, waiting for info/meetings. If one is willing to throw away your life by working for Evil Corp then no amount of agile/kanban/waterfall is going to save you from its clutches.
    7
  4. If companies find a candidate who is a perfect fit they should pay them more than competitors will. But the companies are greedy - they want perfect fit employees at low prices (they call these market related salaries or similar nonsense). So they leave, costing the company $1mil a year to save them $100k a year.
    6
  5. Beware the focus on the now that ignores the near future. Translators 5 years ago: AI can't translate complicated documents. Translators today: AI is just better than 90% of professional translators. Each year AI will get better at more complex systems, working how programmers want them to, and refactoring.
    6
  6. The worst I ever saw was that the international head office in Zurich mandated all subsidiaries to use SAP. That would cost us 5 years profits for a small standalone subsidiary, so we just created a small interface into another copy of SAP running at another local subsidiary and claimed we were on SAP, and that kept them off the trail for the four years I worked there.
    5
  7. Large inflexible engineering is usually created by large inflexible organizations. On the other hand, there are those who with 10x less resources beat out all competitors - SpaceX, Cray in the above video, Google search and so on because they are agile. If you are an inflexible organization you are going to injure yourself trying to do flexible things ( see the Boeing 737 Max Fiasco) - but conversely don't cry when you become the next Kodak, Blockbuster, Yahoo or IBM because somebody else was flexible.
    4
  8. Humans tend to be unable to understand older code as it gets larger. I remember the Boeing 777 had a dangerous bug somewhere in its 200000 lines (if I recall correctly) and it took them years to find it and delayed the release of the plane - costing them serious $. Somebody where you worked should have abstracted the SP into smaller layers long before this point. Of course, they didn't.
    4
  9. The issue with this is that it assumes you know what the hard part is which assumes you have deep knowledge of the problem domain. In the real World, you write what you THINK is the hard part first, then tell the boss it is all going to be done on budget. A month later you discover the REAL hard part and you are up a creek without a paddle.
    3
  10. It might also be worth adding that you go through stages in your life and what appeals to you at one stage is often not the same at the next stage.
    3
  11. Alpaca is not trained on your laptop - it is fine tuned on your laptop. There is a big difference.
    3
  12. Elon Musk: AI is an existential threat for humanity. Elon Musk one year later: I built the biggest AI machine the World has ever seen... Elon Musk 10 years later: All your base are belong to us.
    3
  13. Eagle use was highly risky because of Nazgul on flying beasts. Also, the characters do not think in such straight forward fashion, except for Boromir, and we know how that story ends.
    3
  14. Also, the nature of Sauron was he foresaw all obvious risks, only the most unlikely approaches had a chance of success.
    3
  15. It could be argued most drivers don't know how cars work and they couldn't be helped reading books about engines. Sorry, M8, couldn't resist.
    2
  16. No. Mistakes never result in jail. Intent and recklessness are what send you there.
    2
  17. ding ding
    2
  18. In 1886 when Benz created the first car they were very few who though it better than a horse or walking - after all it was likely to explode and kill the driver, or just fall over. It took 60 years for it to eventually replace the horse as the preferred method of transport. The point I am making is that the more we embed intelligent things into machines the more we will find it difficult to compete with them - alrady over the last five years Google Translate constant improvements killed translation as a career (outside of work for governments and spy agencies), the same is happening right now for artists. Programmers may be better positioned to compete or they may simply be deluded by version 0.1 of something that will improve rapidly from here - in two or three years we will know more.
    2
  19. I studied data science to completion before moving to software engineering. To be fair, these concepts and traps were brought up many times in the subject matter - remember, most of the lecturers are battle hardened software engineers before becoming data scientists. The question therefore remains why are they struggling to do this in practice? Part of the problem is size of the datasets- part of the problem was the Gold Rush in this area 2016-2019, part of the problem is the individuals concerned, part of the problem is Dave's experience is not representative of what every data department is doing. The field of Data Engineering exists to solve this problem and firms that hire these people will naturally avoid these problems. All that said - every point in the video is golden.
    2
  20. More please.
    2
  21. In US politics there is Democracy, but no consent = ever escalating feelings of ill will.
    2
  22. Bad execution is like bad ingredients, they cannot be used to dismiss the recipe.
    2
  23. "developers to compete" Here is the problem. If you hire competitive types they will spent their efforts destroying their colleagues and then jump ship to your competitor for 10% pay increase. Rather hire people with a collaborative mindset. Aside: Read about the experiment when scientists wanted to breed the best chickens by selecting only the strongest most competitive chickens. End result was chickens who just killed other chickens on sight and battery productivity dropped to zero.
    2
  24. 'Canary deployments are not always possible, either for technical reasons (users must be networked) or for business reasons (the want the benefits now). That said, I was the manager of a business where just before I started they went live with a system that worked great in development and failed big in production, mainly because the damage it was doing was hidden from users for a couple months. The damage to the business was extensive and took about three years to repair reputational damage with customers.
    1
  25. Correct, but some problems are really good at hiding from sight before you start developing or even go live. That only means you should double or triple your effort to find these problems before choosing.
    1
  26. I come from an industry where code of ethics was considered a real big deal. What it did in reality was create a shield for people to use cleverly constructed arguments behind - and billions were lost and people jumped and life went on - and nothing changed because we were 'ethical'. What you really need is whistleblower protection program - so those who want to spill the beans don't suddenly find themselves in Tony Soprano's trunk on the way to the lake.
    1
  27. Five QA's have watched this video.
    1
  28. It depends on what software it is. Put up a scaffolded website with .Net or similar - sure. Program the flight systems for a passenger jet - no ways.
    1
  29. Um no. Most serious data products are built by experienced data scientists, a discipline with practices and principles going back to the pen and paper only era. Much care is taken to organize the work and the assumption that error is present is the guiding principle.
    1
  30. GPT pushed an update yesterday or so with new maths in it. Elementary maths is now much more reliable based on my limited testing. Bigger numbers still bugged. There are other AI made that can do very accurate maths so in time they will fix the maths problem as well.
    1
  31. The idea that neural networks are impossible to predict is antiquated and comes from their early years when proponents of older AI models tried to use the same techniques from those networks to understand neural networks. Techniques such as latent space or simply building a smaller, human understood test set, as well as data science/statistical techniques - these can all give a very good grip on a neural network's functioning, at least a good a grip as, for example, Microsoft has on what it has written in the Windows source code, which is not perfect.
    1
  32. He intentionally wrote bad code to mislead regulators. He knew exactly what he was doing. He wasn't he only one involved but if you make guns and give them to people you know will murder people - you share in the guilt - this is part of the law of all countries.
    1
  33. Don't get me wrong - but he makes the reasons for that clear.
    1
  34. I think he means the dataset on which the notebook was based has not been kept as a copy - maybe it is 22TB etc. Of course the notebooks themselves are backed up.
    1
  35. The movie A Judgment at Nuremburg is probably the best examination of the concept of following orders vs complicity. Warning - hard hitting and viewpoint challenging.
    1
  36. GTP3's randomness is not part of the model - it is added using regular code like C++ post-training. This is to prevent it being boring. They force it come up with things to say that are far from the most probable, just to keep you entertained. Tesla does NOT do this with its self driving AI for obvious reasons.
    1
  37. 5:40 Randomness: This is generally speaking, not correct. During inference, the same result is generated from the same input in traditional networks without any random interference other than noise coming from the data itself. What you are referring to is likely things like DALLE which are trained to turn random data into a text prompt - but this is far from the standard type of neural network used.
    1
  38. Aside: There is a random starting point when training data. But in 99.9% of cases if you retrain that network it will form via another route into something that behaves exactly the same as neural networks are exceptionally good at getting every last bit of pattern matching out of the data available.
    1
  39. Regarding intelligence of AI, modern AI is only one type of intelligence - pattern matching, combined with logic and computation types through traditional programming. Scientists reckon there is over 20 distinct forms of human intelligence (e.g. understanding World state, intuitive reasoning, good decisions from small data ) which AI is just terrible at. This it is not a competitor to human intelligence in terms of breadth, but in terms of its sheer power in pattern matching it has us beat hands down.
    1
  40. If you look at set theory 'part of' should be defined as either an intersection, union or subset. Just saying "it is a part of x" does not fully describe the relationship.
    1
  41. That is fine when the tool choice does not matter so much, but if everything hinges on the success of the project and its tool choice - then gambling for fun may have harsh consequences.
    1
  42. Test hypothesis or eliminate hypothesis?
    1
  43. Thanks for sharing.
    1
  44.  @1foxmark  Boeing 737 was because they wanted to copy Airbus wing tips but it required changes the plane was never meant to have. Was not faulty code - was faulty management.
    1
  45. Cables can fail. If statement can fail. It is hard to make the call which is safer. I do know that Teslas with self driving are like 10x safer than regular cars - going by deaths per km. Software is not necessarily more dangerous than hardware systems.
    1
  46. I often get the "I don't know that stuff " answer from ChatGPT so the Dunning Kruger criticism is not a fundamental limitation - merely a problem with not enough or inaccurate training data - meaning it will improve over time.
    1
  47. 15:00 This idea is incorrect. The 30000 pieces of a Coronavirus can bend and curve in near infinite ways, and those curves and bends have an absolute effect on the functioning of the virus. Top genetic researchers will also tell you that genetic code only account for a small percentage of what is going on with biology - they are missing a huge piece of the puzzle (maybe some additional smaller structure?) - the idea that life is just proteins printed off a gene blueprint was a convenient fiction propagated decades before by speculating scientists - actual research paints a different picture.
    1
  48. None of that matters. The courts will look through things like that to the realty of what is going on. At the Nuremburg trials the Nazi regime all claimed they were following orders and so it wasn't their fault. They were found guilty because there is a principle in law that if you knowingly assist in ANY WAY to commit a crime you must share in the guilt.
    1
  49. If corporations were run with common sense and common decency 80% of people would not hate their jobs.
    1
  50. The developers or the management who set the budget?
    1