Trend(s) to Watch: Generative AI

Insights on artificial intelligence, generative AI applications, and next-gen edge services

With a new year upon us, it always presents a great opportunity to look at emerging tech trends. We spoke with Arda Ozgun, CEO at Edge Signal and Vice President, Product Management at Wesley Clover International, to gain insights into the trends shaping our industry. With 30+ years of experience in transforming business ideas into products and go-to-market strategies, Arda brings a wealth of knowledge to many of our portfolio companies.

Arda, to start, what important tech trends are you observing that we should all take note of?

Typically, there are 5-10 key technologies trending each year. However, this is an exceptional year. I have worked in tech for many, many years, and I have never seen one technology overshadowing tech trends on all domains: Artificial Intelligence (AI). Just look at Google Search trends, hacker news, and new startups that are being created, AI is making headlines worldwide. AI chip companies cannot keep up with demand, and that imbalance is expected to continue well into this year.

Artificial intelligence, especially generative AI, is the most revolutionary technology we have seen in decades, on par with computers, cellphones, and the internet.

What is generative AI, and how does it differ from earlier AI techniques?

As humans, we consume vision, text, and voice. That’s what generative AI consumes as input, and as output, it produces a variety of novel content, such as images, video, music, speech, text, software code, and product designs. It’s very human-like and that’s exactly what drives its popularity.

Generative AI is also different from earlier AI techniques in that it is capable of generating new content based on past content. Generative AI models learn the patterns and structure of their input training data and then generate new data that has similar characteristics. With this, everyone can understand generative AI, see its value, and – most importantly – use it.

Most people I talk to associate generative AI with ChatGPT, which – as we all know – produces content that resembles human-generated work, like news articles, poetry, or computer code. It’s impressive but there is so much more to generative AI. For example, at Edge Signal, we are in the process of deploying generative AI in retail shops to provide our customers with valuable insights to understand store dynamics, such as shopper insights, movement and queue analytics, and even customer sentiments.

The main difference between generative AI and other AI techniques – machine learning, as an example – is the human element. Machine learning provides a lot of value for pattern recognition, summarization, anomaly detection, and similar use cases, but those are more specific and not everyone can easily apply them and understand their value.

In your opinion, are there any risks or concerns associated with generative AI?

Yes, most definitely. Huge opportunities tend to bring about risks and many questions that need to be addressed. The regulation of artificial intelligence is very topical in Europe, of course, not to mention privacy concerns. Canadian and American governments are actively exploring where and how to apply policies as well. Generative AI can be misused, including cybercrime, the creation of fake news, or the production of deepfakes that can be used to deceive or manipulate people.

False negatives and false positives will also be of concern. Since you receive messages in a human way, you tend to believe them, but generative AI data is not necessarily 100% correct.

Insights that are biased or not explainable are another issue. For example, AI can be extremely powerful when it comes to healthcare by monitoring patients remotely 24×7 and using machine learning algorithms for pattern recognition and to inform treatment. However, what if AI-generated data suggests that a patient’s condition is worsening, but the patient states the opposite? Inaccurate or biased results could make human validation essential – potentially offsetting its benefits or leading to problematic situations.

Biases could also be introduced into AI systems in many ways, such as biased data, algorithms, and biased human decisions. These biases can lead to unfair treatment of certain groups of people, and can have serious consequences in services, for example around hiring, policing or financial services. Hence, the technology is not ready yet to be used in mission critical or sensitive scenes, but it is a great co-pilot to us.

What comes next?

I predict that we will see a lot of change in the market by large organizations bundling AI into their existing products and increasing their overall value. On the other side, smaller players will try to make better versions of big companies’ products by unbundling valuable portions. From early applications including code assistance, search, autosuggest, and brainstorming, generative AI will evolve from generic to very specific use cases.

AI will further improve labor productivity as well. Jobs are being optimized and mundane tasks are made redundant or automated. Co-pilots will help professionals perform their tasks more effectively, and the same applies to students, social workers, and others.

Also, the shop automation I mentioned earlier does not require a huge server farm if the use case is focused enough. We train the AI for the shops specifically and run it at the edge. This is one example of a very specific AI use case, and I foresee many others.

You mentioned the edge. What role does it play when it comes to AI applications?

Edge is very important. A lot of digitization is done in centralized systems but not at the edge – yet. All centralized data is available for AI today, but not the edge data. What is happening on the manufacturing floor? What real-time insights can be gained from video camera feeds?

This is the next big jump for AI, to process this data that is coming from the edge. If you think about it, most data in the world is generated at the edge, not in centralized systems where there is less sensitivity to latency and business criticality.

Edge has two sides, one is to gather the data, the second is edge intelligence – to unlock many untapped use cases. AI models are making the edge more intelligent, and you don’t need huge servers to run those very specific use cases.

To wrap up, what advice would you give others when it comes to AI and generative AI?

The biggest takeaway is that everyone should start leveraging AI in general and generative AI in order to not be left behind. There is so much more to generative AI than just ChatGPT. Companies need to understand that, apply it to their own vertical, and use it to their advantage.

It can seem daunting to implement generative AI projects. This is why we built an AI-powered edge computing platform, Edge Signal, to enable our customers to use AI resources more efficiently by cleansing and filtering data, and running preliminary AI algorithms at the edge. Edge Signal helps pinpointing anomalies, analyzing visuals, providing insights, monitoring customer behaviour, driving predictive maintenance, automating local machines, and more — leaving businesses to focus on their operations instead of the intricacies of edge computing or AI.

Data collection at the edge is important to support the growing need of training data for generative AI models. With the help of purpose-built generative AI, large language models (LLMs), and reduced costs of AI accelerators, there will be a big potential to run those LLMs at the edge in the future.

To learn more about edge computing and the Edge Signal platform, visit edgesignal.io.

 

Wesley Clover invests in a range of technology companies, and they bring impressive innovation to markets and clients around the globe. I/O is our way of sharing some of the best insights. I trust you will enjoy them.

Terry Matthews, Chairman