Beep boop - this is a robot. A new show has been posted to TWiT…
What are your thoughts about today’s show? We’d love to hear from you!
The problem with AI is its energy usage and consistency. Someone on TWiT at the weekend summed it up, with the new Assistant AIs from Google and Amazon, they answer some questions well, but they aren’t consistent and when it comes to the smarthome, they can’t do the same task the same way every time. Sometimes you need the “dumb” part of the old assistant to carry out the mundane tasks that have to be done one way - like setting a thermostat or turning a light on or off. You might need AI to interpret the human voice, but you just need a simple program to flip a virtual switch or turn a virtual dial, you don’t need a generative AI to be creative.
The same is true of work, the ERP example that Richard mentioned. There is a “simple” (relatively, in terms of the complexity of AI) and a structured database that store the information and display it in a consistent form to the user. It is, compared to AI, relatively efficient and comparatively very energy friendly. The results it outputs are displayed in consistent patterns.
The users know where to look on each form to find the information they need in the order they need it. Getting an AI to do that makes the task less consistent, less user friendly (it won’t necessarily always get the right information (hallucinations) and won’t necessarily display it in a consistent form, or maybe not structured at all, making it harder for the user to do their job, and all of that while at the same time using much more energy to accomplish the task. I really don’t see the benefit for 90% of the work at the moment.
If you need to call up the delivery note and the material safety sheet for the driver, they are standard documents that always has to be pulled up the same way and no variation is allowed - heck, the safety sheets are generated once when the material is developed in the lab and again when ingredients change and the PDF is just printed out, along with the delivery note that is generated from the ERP software, because it would be too time consuming for the ERP software to generate the sheet on the fly every time.
The same goes for the ERP system, 95% of the functionality works better and more efficiently without AI. That said, we are testing AI to automatically read in delivery notes for good being delivered, so that the information can be extracted from the forms in different languages, then the resulting information displayed in a common form to the user to double check, before it is automatically posted to the ERP system.
These sorts of tasks are where AI makes sense in the ERP system, but replacing the whole ERP system with prompts? No, the users have to work quickly and using the simple patterns of information they already have is more efficient, for the computer and the operator, than an AI vibing away. Even if the AI could produce the same layout every time, to make the users’ lives easier, it would still be less energy efficient, by a large margin, than how we are doing it now.
I think AI can enhance current systems and some may be replaced, but a lot of things are so much more efficient than an AI and the AI can’t really bring much more to those tasks, other than making them more expensive to accomplish.
It is the same as the computer in general, we have to look at where and how they can make a positive difference. That is why there are still so many manual tasks, because a computer can’t replace every step of a process efficiently and cost effectively. AI will be the same, it can make general queries for information about large stores of data easier (reporting out of a data warehouse, for example), but the actual usage of the underlying systems and the entry of much of that data will still be quicker, simpler, cheaper and more reliable using traditional programs and operators. AI may support them in some areas, like the scanning of documents and extracting the information, if it can do it 100% reliably, but I really don’t see it, at the moment, replacing all traditional software. It isn’t reliable enough and companies are looking at cutting costs, not making existing processes 10x more expensive to accomplish.
1 Like