Machine learning in 2024

Expected GenAI developments in the year ahead relating to graphics performance.

Ross Cunniff

We will remember 2023 as the year of generative AI (GenAI). For businesses, it was mostly about trying to understand all the possible use cases, both for operations and in their products. 2024 will be the year when companies put into production many of the ideas they are currently considering.

In the area of graphics performance, the biggest trend will be a tighter integration between computer graphics displays and machine learning (ML) throughout the computer graphics pipeline, from the creating and curating of content, all the way down to how pixels are shown on a screen. With that in mind, we anticipate four key developments in 2024 related to graphics technology performance.

1. Democratization of content creation.

Many individuals are already using GenAI to support research, draft documents, and create images. In 2024, we will see a much broader democratization of creativity, enabling individuals who lack expertise to do more on their own.

For a simple, practical example, let’s say you’ve decided to remodel your kitchen. You likely have a general idea of what you want your kitchen to look like—the style of cabinets, countertops, flooring, etc.—but you really need to go to a professional designer to ensure you’ll achieve the overall look you want and the usability you require. By the end of 2024, you may well be able to take some photographs of your current space, enter some dimensions and parameters (such as “an open floor plan with a movable island” or “maximize counter space”) and have an application lay out a design for you. You’ll be able to experiment (endlessly) by swapping specific features, such as butcher-block instead of granite countertops, or wood veneer vinyl flooring instead of beige ceramic tiles. An app may even soon be able to produce drawings you can use to get a building permit, specifications you can use for shopping, and a list of stores where you can get the best deals.

GenAI-powered design isn’t just for the untrained. Expert designers, especially design engineers, will be able to explore design concepts that were previously too time-consuming to develop or too complex to manufacture. In 2024, the combination of ML and additive manufacturing will enable far more creativity and experimentation with many more parameters than is currently possible. Today, for example, automotive engine designers must start with many givens—inline six cylinders, location of the fuel injection ports, and so forth. By the end of next year, they may be able to start designing based on specific goals: What are some options for hitting a specific point on the performance efficiency curve? How can we optimize the use of the non-occupied space to enable more space for batteries?

2.  The democratization of creativity will lead to broad demand for guardrails.

Fear around the consequences of GenAI is already prevalent, and 2024 will see both an industry-wide and consumer-led focus on the ethical use of the technology. GenAI will make it significantly easier to create deep fakes of world leaders, politicians, celebrities, or anyone else who ends up in the crosshairs of someone with an ax to grind. How can we ensure that images of people are not used without consent? In the US, how can we stop abuse without trampling on the First Amendment? How can intellectual property, such as copyright and trademarks, be protected, while allowing fair use for creative purposes? And who owns the intellectual property of GenAI content derived from publicly visible data?

Tech vendors, enterprises, government bodies, and other concerned organizations and individuals will certainly begin formulating strategies and guardrails to allow for innovation, while protecting privacy and limiting abuse. It won’t be easy, and it certainly won’t be resolved in 2024, but the effort will be there, and we all have a stake in the outcome.

3. Better apps and more powerful devices are coming.

In 2024, software vendors will begin using ML-assisted programming to find more efficient and effective ways to organize and display applications and data, leading to enhanced productivity for both the developers as well as the end users. Likewise, at the silicon level, ML algorithms will dramatically change chip design, with static rules-based layouts quickly giving way to more dynamic, ML-assisted layouts that deliver more efficient use of the floor plan. As a result, expect to see announcements related to the increased power and efficiency of desktops, laptops, phones and other devices without sacrificing performance.

4. GenAI will help solve GenAI’s biggest problems.

Ironically, GenAI will be part of the solution to reducing GenAI abuse by helping to identify deep fakes and other misuses. ML will also provide important tools to improve the quality of research. For example, we have recently seen multiple instances of fraudulent research, for which researchers manipulate results, present fake evidence, and even dishonestly use ChatGPT. ML can help to identify many more instances of this, and far more quickly, because the model has access to a corpus of knowledge that people couldn’t acquire in a lifetime.

Another example is that because GenAI is based on analyzing massive amounts of data, often in real time or near-real time, it requires a tremendous amount of processing power and consumes a tremendous amount of energy. However, we also have a responsibility to ensure we don’t worsen the climate crisis. We may not achieve the balance we need in 2024, but researchers are dedicated to understanding how to leverage AI and neural networks to achieve ever greater degrees of power efficiency.

Imagine That

The single biggest impact of GenAI is that it will enable activities and workflows that we can’t yet imagine—and, therefore, even our most fantastic predictions for 2024 and beyond may fall short of reality. I don’t think it’s an exaggeration to say that GenAI will have as profound an impact on society as the printing press!

Ross Cunniff is the chair of the Standard Performance Evaluation Corporation’s (SPEC’s) Graphics Performance Characterization committee. He has more than 35 years of experience in the tech industry, including 25 years with Nvidia, where he serves as a systems software development manager.