Voice of the DBA

Tag: AI

  • Digging into an AI Algorithm

    One of the more successful uses of AI (artificial intelligence) has been in the medical field. It seems that there is a tremendous amount of data, high variability in some aspects of the target of the data (the patient), and a need to constantly review very tedious volumes of data. A place where AI might, and has, made a difference to help humans, not replace them.

    There’s a neat article that dives into a bit of how a medical algorithm works for detecting sepsis. As you scroll, the article tries to explain in layman’s terms what the AI system(s) are doing. It goes through the necessity of regular data, which I think is a good design to push humans to gather data. It then talks about a variety of neural networks being used to analyze data, based on previous training from humans and past results.

    It’s interesting to think of multiple nodes coming to their own conclusions are different times of the day and then later nodes looking at the results of previous nodes. That’s what a human would often do, looking back at recent history and giving that some weight. The system comes up with a number that a human can use to consider in their diagnosis.

    It is interesting that the article talks about the need for humans to better communicate to use this system. I think that’s often a key with any computer assistance. The humans still need to use the computer as a tool and not as the final decision maker. However, at least in this instance, the tool seems to be helping reduce deaths in the hospital. I would hope that this is because the algorithm is detecting some patterns that are easy for a doctor to miss. It could be because the new system is focusing people’s attention better, and possibly this is a combination of both of these possibilities.

    I do think that AI systems can be very helpful in assisting humans in many tasks, and medicine might be one of the places where AI will be most useful. The sheer volumes of data, the complexity of the cases, and the load placed on medical workers to monitor many people can overwhelm any one doctor or nurse. Having a system that might remind them of a small detail will hopefully save lives.

    Steve Jones

    Listen to the podcast at Libsyn, Stitcher, Spotify, or iTunes.

  • AI Ethics

    I found this article to be an interesting look at how we might add ethics to AI systems in one area. As the article points out, “… today there is no broadly accepted AI ethics framework, or means to enforce it. Clearly, ethical AI is a broad topic …”.

    Glad someone is thinking, or many people are, but sad that we aren’t really moving in a direction that creates a better system for us humans to work under, or be bound by. To be fair, I do think this is a very difficult topic and hard for any large group of people to agree on what should be done.

    Network monitoring is a fairly narrow problem domain, at least compared to many others. The article notes there are places where AI can, and does help humans that work in computer networking. That being said, how does the AI handle ethical considerations. For example, can we ensure the AI handles data privacy appropriately. This could be in compliance with some regulation like GDPR. It could also be in a manner that doesn’t disclose data inside a company to other systems or humans who shouldn’t see the specifics of network traffic (like passwords, credit cards, or any sensitive information).

    There are also other considerations. While we see bias in AI systems trained on previous human behaviors, because humans are biased, will network AIs similarly have bias? Will they be less helpful for power users, who have a wide variety of traffic? Those are often privileged users, who might benefit the most from helpful monitoring. Will AIs discriminate against a user when another humans trains or influences it against them? A crude example might be a network admin that doesn’t like women. They enforce more strict rules against women, and influence the AI to do the same. How will the AI, or others, detect this type of issue?

    Maybe the most difficult thing with AI is with corner cases. The ethical dilemmas that might not be easy to solve can confound an AI. Maybe the most ethical choice here is to seek other counsel or let other people help make a decision. To be fair, this is hard for humans to do as well, but for some reason we seem to trust computers less. At least some of us do.

    Ethics is a challenging issue. I find it to be difficult as a human, and my own inconsistency means I might react differently at different times or in different places. How we translate that to AI systems, which are increasingly a part of our world, is going to be hard. I don’t have answers, but I lean towards transparency, accountability from the humans in charge, and the ability to reverse (and apologize for), poor decisions.

    Steve Jones

    Listen to the podcast at Libsyn, Stitcher, Spotify, or iTunes.

  • A Strange AI Achievement

    We are seeing AI and ML becoming used in more and more industries, but one that seems to be a place where it is embraced with some success. From speech recognition and transcription to analyzing imaging, computers have helped medical professionals improve the care they give to patients.

    Just as CAD has helped manufacturers, AI systems are being used in medical research, trying to model and screen medications to try and determine which ones are potentially useful in treating various diseases. We also have used them to better tailor treatments for certain diseases, like some cancers. It seems that applying computing against vast troves of data is proving itself beneficial.

    What if an AI were able to develop something new and it were awarded the Nobel Prize? That’s the premise in a scenario published in the Economist. The article opens with the controversy of how the guidelines for the prize are interpreted, paving the way to award the prize to the AI. Perhaps even more interesting in the scenario is that the effort is the result of a poor software upgrade that allowed the AI to read more medical papers that it was previous given access to examine.

    It’s an interesting idea. Who gets the credit? Certainly, the humans that help train the model and put it to work deserve some credit, but they are really the assistants. If they were to use the prize for more research, do we think they could replicate the innovation? Maybe it doesn’t matter. Not many people win the Nobel prize twice. Perhaps using the prize to continue allowing the AI to conduct research would make the most sense, though I don’t know if the humans around the system would accept that.

    I don’t think AI systems are more intelligent than humans, but they can consider and try more possibilities than humans, given enough data. They might notice something that we’d miss, and they can remain more focused on a problem than we can. After all, we need to rest and care for our bodies.

    I don’t know if this will happen in the 2030s, as the scenario imagines, but I do think this is a possibility as we start to use computing to search for new innovations in research. I just hope that as these discoveries take place, they are used to better the entire world, and not just enrich a few humans.

    Steve Jones

    Listen to the podcast at Libsyn, Stitcher, Spotify, or iTunes.

  • Finding Legal Data

    Many of us have likely been asked about data science or machine learning in the last few years by someone in our organization. This has become a hot field, with many companies looking to try and find ways to use the technology to improve their work in some way. While I don’t know how successful these projects have been in organizations, I know that some areas seem to be finding practical applications. Image recognition, translation, and even some application scoring systems have benefited from machine learning algorithms.

    To build a successful model that works well, we need training data. Often lots of training data, and then have some metadata about how our training data might be applicable to a particular question. For example, if we have lots of pictures of dogs, we might need to tag the different breeds in order for a system to differentiate among them. If we want loan applications scored, we should have a corpus of documents that are already scored. This metadata allows the system to learn.

    There is a company, Clearview AI, that markets itself as a facial recognition system. To build their model, they scraped images from YouTube and other Internet sources, without consent from Google or the subjects of the videos. This is interesting, as the data itself is publicly available, but gathering it into a database and using it for other purposes might run afoul of data privacy laws, like the GDPR.

    I don’t quite know how to feel about this use of public data. While I don’t mind people viewing my pictures, I’m not sure that I like the idea of them being copied into a database for some other purpose. That might seem silly, or even strange, but I do think there is power in data and more power in more data. Allowing others to put my images in a database and use them, perhaps to train a model to recognize me, feels like overreach.

    If you need data for your company, or your idea, what can you do? Many people just scrape Google, Facebook, etc., and get data. That might cause you some legal issues in some places, and you ought to be aware of the implications if you choose to do this, or you are asked to do this. I don’t think this is how we want data to be gathered. I know there are some guidelines for responsible AI at Microsoft, but not necessarily rules in place for many companies. Hopefully that will change over time.

    Steve Jones

    Listen to the podcast at Libsyn, Stitcher, Spotify, or iTunes.