Commvault Unveils Clumio Backtrack - Near Instant Dataset Recovery in S3

// 24 Oct 2023

This is not the AI you are looking for: Reflecting on Gartner IT Symposium

Jacob Berry, Field CISO
ShareTwitterfacebookLinkedin

A healthy dose of skepticism

Those of us who have held technical roles often are skeptical. Skeptical of new trends, new buzz words, and vendors.

For many of us it’s rooted in spending many a late night grinding to get technology to work, or more specifically, trying to get technology to integrate and provide meaningful outcomes. It’s hard to understand what will work or won’t work ahead of production, even if we can afford full-scale testing and proof of concepts (which most can’t).

All of us have had tech fail, missed timelines, or blown budgets.

We feel we need healthy skepticism to aid in our success.

As the Gartner IT Symposium conference unfolded I talked with peers and technology leaders about the sessions and what they were hoping to take away. I can summarize the thoughts in these questions:

  • What should I be skeptical of and what should I embrace when it comes to AI?
  • What should my stance and AI usage policy be for my organization?
  • Is there any AI technology that can be adopted now and what’s the impact on my business?

Leaders entered the conference with uncertainty about the buzz of artificial intelligence and hoped that there might be answers.

Cutting out the noise

Before we find out if these questions were answered, or if we left just as perplexed as when we started the conference, I feel I need to add some clarity to the way I think about AI.

There is a lot of noise both in legitimate new tech, products adopting “AI,” as well as a fair smattering of people riding the wave.

I don’t think AI exists. Not yet at least.

I say that as someone who has an inclination towards science fiction. SciFi has defined AI for a long time, typically as something that can act as an independent agent with the same level of intelligence as an average human, but is not an organic being.

Unless you’re reading this on your commute back from a secret underground lab where this exists, we don’t have anything that can act truly in this manner.

So, if by that definition AI doesn’t exist, why is “AI” splattered on every banner and headline?

Because we have steps towards this definition of AI. Meaningful steps.

And AI is an easy label to apply. For the average person (as in someone who doesn’t have a masters or doctorate in computer science) AI catches attention and is digestible. The nuance of the individual technologies is hard to understand. But once we pull back the generic AI label we can have a meaningful discussion.

There are roughly four sub-categories of technologies that fall under the category of AI:

  1. NLP – Natural Language Processing. Using a multitude of machine learning (ML) algorithms, or algorithms that essentially can create accurate outputs from inputs without the developer having to explicitly map inputs to outputs, we can infer certain information from human written text such as overall meaning of semantic similarity of bodies of text. NLP can be used to take natural human text as an input without having to code every single condition into an application.
  2. Generative AI / ML (Gen AI) – This is really what the buzz is about. This also uses many ML algorithms to not only understand input from human text, but can also interpret images and audio in a way similar to humans and then create novel outputs in text, images, and audio. ChatGPT, Bard, and others are examples of this.
  3. General ML application – More broadly, often general machine learning algorithms (as defined above) have also been referred to as “AI” – typically a specific use of machine learning is focused on understanding large volumes of data and being able to find patterns, create accurate predictions, create summaries, or use large volumes of data to complete tasks. Here is a really cool example from a few weeks ago that is tangible: Training an unbeatable AI in Trackmania
  4. Predictive analytics – Lastly, predictive analytics and data science tasks (Both ML and Statistics) have often been labeled as AI, although they are furthest from my definition above. However, predictive analytics has one of the most meaningful applications. For example, in manufacturing, predictive maintenance and interpreting hundreds of signals in a mechanical system can predict when parts need to be replaced, preventing outages and thereby increasing manufacturing uptime.

With this in mind I think we can get back to discussing my takeaways from the conference.

What I took away from the conference

Let’s look at the takeaways in two categories: Tips for your AI policies, and where using AI technology makes sense.

Most folks are concerned about three root issues:

  • Can publicly accessible Generative AI tools expose my sensitive intellectual property if the IP is used as input in the tool? (Unfortunately it’s not a yes or no answer)
  • Can Generative AI cause employees to rely on incorrect information and create systemic issues for my business ?
  • What from Generative AI should I adopt so my competition doesn’t gain a “counter positioning” power in my market?

I think, based on my conversations at least, this is what the true concerns are. I didn’t include copyright based concerns as I think it’s a more generalized conversation about generative products, how they are created, and what Copyright means in a post generative AI world.

Here’s the body of the takeaways from the conference.

Tips for your AI Policy

  • Be specific. You likely need a usage policy to address the risks associated with generative language tools, and perhaps generative image or audio tech.
    • Specify which technologies you’re addressing, how they should be used, and provide examples.
    • Focus on addressing data loss, IP loss, and copyright risk.
  • Don’t ban usage.
    • Humans are lazy. Generative language tools are going to be used by many, no matter how you try to stop it or ban it. All you will do is create a culture in which people hide their usage and likely introduce more risk.
  • Focus on education and implementation.
    • Teach people what is appropriate to enter into LLM generative language tools and what is not.
    • Teach people the benefits and drawbacks of generative AI.

Which is a good segway into the next topic, AI usage.

AI Usage

Boston Consulting Group’s report is an excellent analysis of everyday adoption of Generative AI tools in the workplace. There are meaningful ways to improve productivity, but it requires analysis to determine what areas of your business can be positively affected by a generative AI tool.

Generative AI has helped me increase productivity and I have been introducing techniques to my team to increase productivity, but it is not the end-all be-all. You need to know how to use it and when not to. If you rely on generative AI too heavily you will find out its faults when the consequences are highest.

In general it has been effective for me in three ways: 1) organizing my thoughts, 2) helping brainstorm, 3) practicing and coaching various skills.

I’ve also found that for non-debated, well-discussed, general knowledge it’s a good alternative to searching references too – but as soon as the knowledge is niche or requires subject matter expertise, generative tools quickly show their limits. Unfortunately it takes a subject matter expert to separate truth from hallucination in these cases.

Outside general employee productivity, there are many uses that were talked about at the conference that are worth looking into. I’ll summarize some quickly:

  • Helpdesk and support interaction
  • General product Q/A support for finding information faster
  • Better predictive analytics

I think for me the takeaway not only from Gartner, but in general, is that generative language tools are changing how we interface with computers and how quickly we can access information. Finding applications for that is where the most value will be found right now.

If you are one of the many many people who stopped by and saw us at the Clumio booth this week, we were thrilled to see the excited support from both existing customers and folks who were learning about us for the first time.

For those to whom we still owe follow up, I’m looking forward to the meetings!

About the author

Jacob's background is in Cyber Security and Technology, focused on helping customers build secure cloud operating environments. He has extensive experience in offense and defense security, security operations, and working across multiple verticals in both private and public sectors.