• Visitors can check out the Forum FAQ by clicking this link. You have to register before you can post: click the REGISTER link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. View our Forum Privacy Policy.
  • Want to receive the latest contracting news and advice straight to your inbox? Sign up to the ContractorUK newsletter here. Every sign up will also be entered into a draw to WIN £100 Amazon vouchers!

Demand for AI "Surging"

Collapse
X
  •  
  • Filter
  • Time
  • Show
Clear All
new posts

    #81
    Originally posted by SussexSeagull View Post

    I can conceptually understand how it can derive Unit Tests if it has access to code but if you move further up the V towards things like System and User Acceptance testing then it needs human input. In fact it would require humans who actually know what they want which can be a bit of a rarity in software development at times!
    Tell me about - our product owner is a waste of space.

    But maybe... maybe AI would have more of an idea than him? If I click a menu item in an application I would expect something to open that corresponds with what that menu item says it does. An example, I click File -> Open, I would expect some kind of file chooser dialog box to appear. An AI trained to test applications might also have similar "intuitive" expectations of how things should work. So a bit like when you put an application in front of a completely green user and ask them to try it out and let you know what isn't working or does not seem obvious, this kind of task does seem amenable to exploration by AI.

    But I agree, AI becomes useful when you have a human-in-the-loop who is in charge, and AI doing the donkey work, but also being smart about it and not completely screwing it up.

    Comment


      #82
      Earlier in this thread we discussed reasons why AI work might not be coming through to contractors. But I think things are changing. Anyone else starting to see a much higher proportion of job ads or agents looking for AI people? I certainly am on jobserve feeds.

      Comment


        #83
        Originally posted by willendure View Post

        Tell me about - our product owner is a waste of space.

        But maybe... maybe AI would have more of an idea than him? If I click a menu item in an application I would expect something to open that corresponds with what that menu item says it does. An example, I click File -> Open, I would expect some kind of file chooser dialog box to appear. An AI trained to test applications might also have similar "intuitive" expectations of how things should work. So a bit like when you put an application in front of a completely green user and ask them to try it out and let you know what isn't working or does not seem obvious, this kind of task does seem amenable to exploration by AI.

        But I agree, AI becomes useful when you have a human-in-the-loop who is in charge, and AI doing the donkey work, but also being smart about it and not completely screwing it up.
        Well you always have to train AI, some humans have to get involved at some stage. A well trained AI model can definitely do visual checks, heck even a dumb script which does an action based on mouse location and then checks the intended result with a screenshot is fairly useful (seen that done once and even though it seems medieval, it is some for of automation). And even if the AI isn't good enough to tell you if the actions generate intended content / results, it can check for silly errors, like no action / blank screens etc.

        Comment


          #84
          Recently I made a sketch as an SVG drawing of a system that I am currently implementing. Blocks to represent the UI, backend, database and so on. And some relevant parts of the system drawn inside - a message queue, some APIs. Then numbered arrows linking things together in the sequence they will happen.

          Pasted it into ChatGPT o1, and asked what do you think of this design? Just from the drawing it was able to understand enough about what I was trying to do to describe it accurately, and it made some suggestions on areas that could be improved to make them more efficient or reliable. Nothing too amazing, all text book stuff, and for various reasons not actually useful suggestions, things that I had already considered and dismissed. But still... I was impressed. Quicker and better analysis than I would expect a junior dev to have made.

          Comment


            #85
            Originally posted by dsc View Post
            Well you always have to train AI, some humans have to get involved at some stage.
            True but...

            Part of the reason LLMs have been possible, is because of unsupervised learning. There is masses of text available to train on, and the training is just a game of guess-the-next-word. Masses of data, masses of compute, and humans just wrote the code for the training, not actually human in the loop to conduct the training.

            Datasets for the instruction following training did have to be prepared by hand.

            o1 was trained using some kind of reinforcement learning - the Q* or Q learning thing that was hyped a while back. Reinforcement learning is also fully automatic, no human in the loop to conduct the training. This is how AIs learn to play video games for example. They just play and play with the aim of staying alive for longer, or getting a higher score or whatever metric their performance can be measured against.

            Comment


              #86
              Originally posted by willendure View Post

              True but...

              Part of the reason LLMs have been possible, is because of unsupervised learning. There is masses of text available to train on, and the training is just a game of guess-the-next-word. Masses of data, masses of compute, and humans just wrote the code for the training, not actually human in the loop to conduct the training.

              Datasets for the instruction following training did have to be prepared by hand.

              o1 was trained using some kind of reinforcement learning - the Q* or Q learning thing that was hyped a while back. Reinforcement learning is also fully automatic, no human in the loop to conduct the training. This is how AIs learn to play video games for example. They just play and play with the aim of staying alive for longer, or getting a higher score or whatever metric their performance can be measured against.
              There's still a human in the loop, the text, whatever you feed it, was made by humans. It's not created out of thin air, someone put effort / work into all that data that's available for training. I know the process is unsupervised, but it's not like AI can just make it's own training data blocks and train itself, you have to feed it stuff made by someone. Same principle with testing, you define a "win" scenario by declaring acceptable results and those are human defined.

              Comment


                #87
                Action plan, no less: https://www.bbc.co.uk/news/articles/crr05jykzkxo

                Comment


                  #88
                  So I downloaded the AI Action Plan PDF from gov.uk: https://www.gov.uk/government/public...nment-response

                  And to be true to the cause, I asked an AI to give me a summary of the plan. Here it is:

                  The AI Opportunities Action Plan Government Response outlines the UK's strategy to harness the potential of artificial intelligence (AI) to enhance economic growth, improve public services, and secure a leading global position in AI. Here’s a summary of the key points:
                  Foundations for AI
                  1. AI Infrastructure: A plan to expand AI computing capacity by 20x by 2030, including the establishment of a supercomputing facility and AI Growth Zones to attract private sector investment and build robust AI data centers.
                  2. Energy Needs: Creation of an AI Energy Council to address the energy demands of AI with renewable solutions.
                  3. National Data Library: Development of a secure repository for public sector data to facilitate AI research and innovation.
                  Skills and Regulation
                  1. AI Talent Development: Expansion of scholarships and training programs to address AI skills gaps and enhance diversity in the talent pool.
                  2. Regulation: Establishment of frameworks to promote safe, ethical AI development and use, balancing innovation and safety.
                  AI for Public Services and Economic Growth
                  1. Adoption in Public Services: Implementation of AI in public services using a “Scan, Pilot, Scale” methodology, with AI Champions appointed to facilitate adoption.
                  2. Homegrown AI: Measures to strengthen UK-based AI companies, attract global talent, and support startups to maintain the UK’s competitive edge.
                  Collaborative Efforts
                  1. Private-Public Collaboration: Initiatives to promote AI adoption across sectors, including partnerships, procurement strategies, and sharing of best practices.
                  2. Global Partnerships: Building international collaborations to enhance compute capabilities and share AI advancements.
                  Next Steps


                  The government will continue refining its AI strategy, monitor progress through the AI Opportunities Unit, and integrate these initiatives into broader industrial strategies by 2025.

                  This comprehensive plan underscores the UK’s ambition to become a global AI leader while ensuring ethical, inclusive, and sustainable development.

                  Comment


                    #89
                    Does it say where AI growth zones will be located?
                    ChatGPT said:

                    ChatGPT

                    Yes, the AI Opportunities Action Plan Government Response mentions that the first AI Growth Zone (AIGZ) will be located at Culham, the headquarters of the UK Atomic Energy Authority (UKAEA). This is subject to the agreement of a public-private partnership that delivers local and national benefits. The government also plans to identify and select additional AI Growth Zones by Spring 2025, taking into account factors such as regional growth opportunities, energy requirements, and alignment with local and national strategies.

                    Comment


                      #90
                      Does it indiciate how much will be invested to adopt AI in public services?
                      ChatGPT said:

                      ChatGPT

                      The AI Opportunities Action Plan Government Response does not specify a concrete total investment amount solely for adopting AI in public services. However, it highlights a commitment to the "Scan, Pilot, Scale" approach, which involves:
                      1. Piloting AI projects in public services to assess their feasibility and benefits.
                      2. Scaling successful pilots with senior support and central funding.
                      3. Developing frameworks and infrastructure for AI adoption, including streamlined data access and procurement processes.

                      While the document emphasizes funding strategies and support mechanisms (e.g., central funding for scaling successful pilots), the exact financial allocation for these initiatives appears to be part of broader commitments tied to infrastructure, talent, and regulatory development rather than a distinct, defined figure. Additional details might emerge in the planned Spending Review in Spring 2025.

                      Comment

                      Working...
                      X