The Obvious:
Artificial Intelligence (AI) will shape the conversational and technological landscape of 2024. While AI took center stage in 2023, most of the practical applications fell into two categories: Experimentation & familiarization. This was commonly associated with “AI washing” yesterday’s tools. The third category worth mentioning is Premature-AI-Exfiltration, where we saw some of the biggest names in tech release AI models well before they had value to offer in an attempt to keep up with the Joneses.
In 2024, we’ll see more formalized business adoption of AI models, both off-the-shelf GenAI tools applied to targeted business problems and custom-trained models focused on solving new challenges. For example, existing GenAI models might be deployed to improve technical support and help desk operations by rapidly synthesizing existing knowledge base for front-line staff. Meanwhile, general purpose AI models could be trained to solve specific challenges for industries like oil and gas.
Lessons to Avoid Learning the Hard Way:
1. How your training data, queries, etc. are used. As always, be wary of the data you expose to 3rd parties, and be clear on:
- How they use it.
- How they learn from it.
- How they share it.
- How they secure it.
- What is your recourse if they leak or use it outside of the stated agreements?
- Where you store your data.
2. AI models are only as good as the data they’re trained on.Training sets will be large data pools which must be accessible by the model. You’ll want to think through things like:
- Where the right location to host the data and model is.
- What the data storage and data access costs will be (including costs like cloud egress charges).
- How you’ll secure the AI model and model access (i.e., who can run what queries), and training data.
How secure the training data set behind the model really is. Proof of concepts have already been shown capable of retrieving original training data via standard model queries. If the data you train with has secrecy, privacy, compliance, or other concerns, you’ll want to know how it’s protected in original form by the front-end model.
The Less Obvious:
Edge computing will accelerate and change form.Edge computing is the concept of providing compute capacity closer to the data source. Originally, the concept was driven by things like real-time analytics.
Example: Using high-definition video feeds to process traffic congestion and optimize for it. This level of data-intensive real-time processing can’t wait for feeds to be sent to the cloud and analyzed before having insights returned.
In 2024 AI will take the driver’s seat in expanding edge compute deployments. This will primarily be driven by AI inference processes, deriving insights from a trained AI model. Whether this is edge resources on the factory floor to optimize processes, or edge resources in hyper-localized co-locations, we’ll see an expansion of Edge deployments.
Lessons to Avoid Learning the Hard Way:
Data center resources have environmental requirements. Edge computing can be thought of as small-scale datacenter infrastructure. After all, you are dealing with compute, networking, storage, and security. This hardware is designed for an air and power condition environment with tight physical security. If your intended deployment location is a manufacturing plant, you’ll need to account for these things. Otherwise, you’ll suffer outages and hardware failures, as well as risk voiding support and warranties.
Not all edge is created equal. It’s safe to assume that the data you process at the edge and the insights derived will need to be accessed by other systems, which is something you’ll want to account for up-front. For remote locations, this may mean deploying public or private 5G connectivity. For many workloads, the best option may be to work with co-location providers that offer enhanced connectivity options. This allows data access from around the globe with fast local cloud on-ramps.
The Prediction:
AI will come together with Artificial Reality (AR), and Virtual Reality (VR) to create enhanced remote experiences. These enhanced experiences will come in many forms, from eye-tracking and correction to create more meaningful eye contact during remote meetings, to full VR-based experiences as promised by the (oft over-hyped) Metaverse concept.
As new tech giants continue to enter the AR/VR space with user devices, we’ll see an uptick in adoption. Subsequently, the amount of use cases will rise as well. Think back to the early days of the first iPhone, before it had a limitless appstore: It was a brilliant device with some proof-of-concept tools and a lot of potential. We will see similar from AR/VR in 2024, in part powered by AI.
Lessons to Avoid Learning the Hard Way:
Tread carefully… Many of these advancements will come in the form of “collaboration tools.” Collaboration is a space that’s already heavily fragmented, and deeply siloed. Customers have to choose a platform, and these platforms make no real attempt to integrate or work together. As companies like Meta and Apple move into the space with AR/VR, we’ll see more fragmentation, not less. This is where the fight for ecosystem integration comes into play. A collaboration experience is only as good as the people you can collaborate with. If everyone has to make the same vendor choice to work together, you’ve purchased a walled garden.
Am I the product, or the customer? For a short period, this was a choice users and companies could make. Now, it is more common you are both the product and the customer. Thus, it’s imperative to know what is being done with the data and metadata collected from tools you use. We’ll also caution you to be cognizant of the potential data-mining risks you’re forcing on your employees by working with companies offering collaboration and Metaverse experiences.
Conclusion:
2024 will see a continuation of the accelerated pace of change and disruption. It will be important for companies to weigh the risks of moving fast against the risks of moving slowly. The odds are that we’ll see more false starts than successful use cases, especially as the landscape shifts rapidly. Opportunity exists in chaos, but it does not come free of risk. The best bet for most companies is to make small strategic bets and adopt a fail-fast mentality for changing those bets.
