AI had a huge year last year, going from a specialized to a mainstream technology almost as quickly as anything has ever done. However, as people begin to acknowledge the potential and constraints of artificial intelligence in general, 2024 will be the year that the hype finally bursts into reality. Here are some scenarios that we believe will occur.
OpenAI turned into a product business.
OpenAI will be a different organization following the November leadership shift; perhaps not visibly, but Sam Altman’s increased authority will have an impact at every level. And having a “ship it” mentality is one way we anticipate that to materialize.
We’ll see it with the GPT store, which was supposed to open in December but was appropriately postponed because of the conflict in the C-suite. Hugging Face and other open-source models will never be tolerated when the “app store for AI” becomes the place to purchase your AI toys and tools. They will adhere to Apple’s model to the letter because it is a great starting point.
Anticipate further actions along those lines from OpenAI in 2024 as an unsightly thirst for markets and customers replaces the prudence and scholarly restraint of the previous board.
I think this tendency will be more noticeable in this situation, but other large firms with AI projects will probably follow suit (for example, expect Gemini/Bard to hone in on a ton of Google products).
Agents produced audio and generated video advances from whimsical to avant-garde
By 2024, a few specialized uses of AI models—like agent-based models and generative multimedia—will have progressed past the “eh” stage.
Artificial intelligence (AI) will require access to your spreadsheets, ticket purchasing interfaces, transportation apps, and other data if it is to assist you in tasks beyond simple summarization and list creation. A few hesitant attempts at this “agent” technique were made in 2023, but none of them gained any traction. We don’t think any will really take off in 2024 either, although agent-based models will demonstrate their worth a bit more than they did in the previous year, and a few key use cases for infamously time-consuming procedures like filing insurance claims will surface.
Audio and video will also find markets in which their drawbacks are less obvious. A lack of photorealism is not an issue in the hands of talented designers, and AI video will be utilized in entertaining and novel ways. Similarly, generative music models will probably find their way into a few big productions, such as video games, where experienced musicians can use the technologies to compose an endless soundtrack.
It becomes evident what monolithic LLMs cannot do.
Large language models have so far inspired a great deal of confidence due to their shown capabilities, which have exceeded everyone’s expectations and have only increased with the addition of more computation. But something will finally give in 2024. Since research is still ongoing at the boundaries of this discipline, it is impossible to anticipate with precision.
In 2024, there will be greater research and understanding about the supposedly miraculous “emergent” talents of LLMs, and things like their incapacity to multiply big numbers will make more sense.
Simultaneously, we’ll witness declining returns on parameter counts, to the extent that while training a 500-billion-parameter model could theoretically yield better outcomes, the computational resources needed to do so could likely be used more efficiently. A collection of smaller, more focused models, as well as perhaps multimodal ones, may prove to be nearly as useful as a single monolithic model, but they are much easier to update piecemeal.
Reality and marketing collide
The plain truth is that it will be extremely difficult for businesses to live up to the anticipation that will develop in 2023. When firms undergo their quarterly and annual inspections, marketing claims made for machine learning systems they implemented to stay ahead of the curve are likely to be found wanting.
Anticipate a significant exodus of customers from AI tools as the advantages do not outweigh the hazards and expenses. At the other end of the scale, we might witness legal actions and government regulations against AI service providers who can’t prove their claims.
The products of 2023 won’t all survive by a long shot, even though capabilities will keep expanding and improving. As the wave’s wobblier riders fall and are consumed, there will be a round of consolidation.
Apple enters the picture
Apple has a well-established practice of waiting, observing, and taking lessons from the mistakes made by other businesses before storming in with an expertly thought-out strategy that puts competitors to shame. Apple’s move into AI is well-timed, not only because its competitors could steal market share if it waits too long, but also because the technology is ready for their kind of advancement.
I would anticipate an AI that leverages Apple’s growing influence in users’ lives to integrate the various signals and ecosystems the firm is aware of, with an emphasis on useful applications of users’ personal data. It will most likely have a multimodal understanding (mostly to manage user photos), but they’ll completely avoid media generation. There will also probably be a deft and elegant method to handle challenging or perhaps harmful prompts. Anticipate some remarkable but highly customized agent capabilities as well: “Siri, book a car to take us and get a table for four at a sushi place downtown around 7.”
It’s difficult to predict if they will market it as a better Siri or as a completely new offering called Apple AI, which you get to name. They might believe that the previous brand is burdened by years of relative incompetence, but since millions of people now say “Hey Siri” once every ten seconds, it’s more probable that they’ll choose to maintain that momentum.
Cases in court rise and fall
In 2023, a substantial number of cases were brought, but very few of them achieved any significant progress, much less success. The majority of lawsuits over copyright and other errors in the AI sector are currently unresolved. Many of them will be abandoned by 2024 as businesses obstruct vital information, such as training materials and procedures, making it harder to substantiate claims like the usage of thousands of copyrighted books in court.
But this was just the start, with more lawsuits brought primarily on the basis of legal theory. Even if they don’t succeed, they might expose enough flaws in the process during testimony and discovery that businesses would prefer to reach a settlement than risk specific facts coming to light. In addition, new cases involving the misuse and abuse of AI will be filed in 2024. These lawsuits will focus on issues including wrongful termination, discrimination in hiring and lending, and other instances where AI is being used carelessly.
However, although a few extreme cases of misuse will be prosecuted, it will only be brought to court randomly due to a lack of pertinent laws. In light of that,
Those who embrace change early grab the reins
Large-scale initiatives, such as the EU’s AI Act, have the potential to transform the sector, but their effects are often gradual. This is intentional, meaning businesses won’t have to quickly adapt to new regulations, but it also means that the effects of these significant laws won’t be seen for some time unless businesses are prepared to make proactive, voluntary changes. There will be a great deal of “we are starting the process of” conversation. (You should also anticipate a few quiet lawsuits contesting certain statutes.)
As a result, we may anticipate a booming new market for AI compliance as the billions invested in the technology drive corresponding (albeit smaller-scale, nonetheless significant) investments in ensuring that the instruments and procedures adhere to national and international standards.
Unfortunately, 2024 is not the year to expect progress on the front of meaningful federal regulation in the United States. The United States government and electorate will be too preoccupied with the political dumpster fire that will be the 2024 election, even though it will be a year for AI and everyone will be clamoring for new laws.
AI exacerbates the trash fire that is the 2024 election.
It’s virtually anybody’s guess as to how the 2024 presidential race will turn out at this point. There are too many moving parts to make any firm predictions, other than the fact that influencers will continue to utilize all available tools, including AI in whatever form is most practical, to sway public opinion.
For example, be prepared for automated blogs and accounts to continuously spew created gibberish. With a text and image generator, a few individuals working full-time can cover a lot of terrain, producing hundreds of fake news and photographs for social media and blogs. “Flooding the zone” has always been a successful strategy, and AI now multiplies labor, enabling more extensive yet focused campaigns. Anticipate both false positives and false negatives as part of a deliberate effort to skew perceptions and cause people to doubt everything they read and see. For politicians who thrive on instability, that’s a win-win situation.
Businesses will use “AI-powered” studies to support voter roll purges, challenges to vote tallies, and other actions meant to impede or stifle legitimate procedures.
We’ll be using both audio and video that have been generated, and while neither is flawless, it’s plausible with some fuzzing: The video doesn’t need to be flawless because it will be shown as a blurry close-up of a hot mic at a private gathering or as a shaky smartphone shot in a dark room, among other scenarios. The question then becomes, “Which of us are you going to believe—me or him?” And for some, that’s all they require.
There will probably be some flimsy attempts to prevent generated content from being utilized in this manner, but companies like Meta and Google are unable to remove these posts quickly enough, and it is improbable that X can or will efficiently monitor and remove such stuff. It will not be a good time!
