What happens when the unstoppable advance of Generative AI meets the unbreakable wall of AI regulations? And what if societal and political issues are heating everything up? Answers to all this and more can be found in Tooploox AI Trends 2024!
James Tiberius Kirk, born March 22, 2233, was one of the first people to communicate with a computer using an activation phrase (“Computer!”) and then having an almost casual conversation with the machine. At least – on screen. And this was one of the top computers of the age, operating all systems aboard the Federation’s Constitution Class Starship, the USS Enterprise. Today, millions of users are getting the same level of service from devices that easily fit in their pockets.
The year 2023 was undoubtedly driven by a ChatGPT-induced revolution sparked in November 2022 when the service was made public by OpenAI. And this was only the beginning. That’s why keeping an eye on trends is so important – not falling behind and surfing the wave as it persists.
What is a trend?
Again, as with every Tooploox article about trends, it is worth writing about what a trend is in itself. The critical aspect to consider is that a trend is broader than a single technology or use case and should instead be considered an umbrella term that imposes order by bundling a group of phenomena together. Considering that ChatGPT cannot be a trend – but generative AI can be and is.
AI trends 2023 review – were we accurate?
About a year ago, the Tooploox team prepared a list of the most influential AI trends for 2023. These were as follows:
- Generative AI
- Trustworthy AI and AI ethics
- AI and society
- Workforce augmentation
- Step-by-step improvements of existing techniques
All major AI-related events of 2023, carefully followed in Tooploox CS and AI news series, show that these predictions were accurate – with generative AI gaining traction as a mainstream technology, the AI act being adopted by the EU, and the building of larger and more effective neural networks.
For this year, the list follows these directions, yet with several twists that have emerged from a year of experience with new technologies and from the facing of challenges unseen before. Who could have said that hallucinations would be feared by the greatest companies and institutions worldwide?
This year’s list of Artificial Intelligence Trends includes:
- Generative AI
- The AI legal situation
- AI’s impact on society
- Hyperautomation and AI integration everywhere
- Environmental impact and chip shortages
Enjoy the read!
Recent Trends in Artificial Intelligence for 2024
In the ever-changing landscape of modern AI and ML, one needs to overcome the fear of missing out when cherry-picking the most important elements from the endless stream of news and information. It is challenging, especially considering the fact that the high-tech industry is prone to buzz and marketing mumbo-jumbo, inflating the meaning of some phenomena and making it easy to overlook the gems.
Yet, with support from the Tooploox research team, all the heavy lifting of scanning through the noise has been done.
AI Trend 1: Generative AI
Generative AI is undoubtedly one of the dominant AI trends of 2023 and 2024, with the potential to become the transformative tech of the decade. “Generative AI” itself is an umbrella term for techniques and technologies that allow the generation of new data built from existing data.
For example, Large Language Models (LLMs) are dominantly used to generate texts from other texts – in longer form or computer code (also text, after all) from a prompt. A prompt is a short text – or instructions – for a model of what must be done. More about prompts can be found in our recent text with 50+ awesome prompts for ChatGPT that every company should use.
Generative AI includes several phenomena to keep an eye on:
Everybody is a creator
According to a popular myth, 10 thousand hours of practice is required to become a master in any skill, be it writing, drawing, playing an instrument, or coding. The accessibility of generative AI tools makes everyone a creator whose output can be comparable with the effect of a professional writer, designer, or musician.
According to data published by Adobe, the Firefly tool has been used to generate over one billion images in the three months since its launch – and this is only one of the popular tools available. ChatGPT-published books have flooded popular self-publishing platforms, with the Guardian warning against reading ones about mushroom picking.
Parents are creating whole books for their children with texts and images – all the work usually done by publishing companies can now be achieved (to a degree, limited by AI tools’ capabilities and the skill of the user) at home.
This phenomenon can be further reinforced by the availability of GPTs trained by users in the GPT market – users can now publish and sell their own GPTs trained for a particular purpose – for example, as email writers or advisors. The number of chatbots designed to help in prompting or writing is also likely to increase.
Video generation
Tech giants have also entered the text-to-video race – for example, Meta has released Emu, and Google has entered the race with VideoPoet. Transforming static images or prompts into films is the next arena where tech giants will compete – and users will undoubtedly find new ways to express themselves. Tooploox Research is also contributing to the development of this area, with Diffused Heads: Diffusion Models Beat GANs on Talking-Face Generation research published during the Winter Conference on Applications of Computer Vision 2024 being only one of the examples of our research work in the field of video generation.
Hallucinations
“Hallucinate” has been named word of the year by dictionary.com for a reason. The tendency of Large Language Models to make up facts that sound plausible is going to be a hot topic in 2024, mostly due to the fact that it limits the tool’s potential. For example, Pak’n’Save’s chatbot, initially designed to provide users with recipes using food they have in their refrigerator, suggested producing chlorine gas.
Boosting open-source LLM’s quality
With the security risks and a lack of true control over the data processed by the LLMs provided by market leaders, companies will turn their attention to open-source models. Having the ability to run a model in-house, train it using owned data, and keep everything within an enclosed environment will likely gain traction.
AI Trend 2: The AI Legal Situation
Changes and challenges in AI’s legal situation will undoubtedly be a trend that will shape the state of AI in 2024. The need to settle an agreement between content creators and Artificial Intelligence companies that use their content to train models is rising swiftly – from the government and private sides alike. And there are multiple approaches.
The German publishing company Axel Springer, has made a deal with OpenAI to enable ChatGPT to access their content and train chatbots using it. On the other hand – the New York Times sued OpenAI for unauthorized use of their texts when training their chatbot. The company stated that it could top “billions of dollars worth of journalism work.”
Regulating the market
The European Union has made one of the first attempts to regulate the AI space, with a goal to avoid situations where malfunctioning AI or biased datasets have an impact on people’s lives. A detailed description of the act can be found in The Artificial Intelligence Act in the EU – what’s likely to change blog post. The act itself was under discussion for a long time, yet by the end of 2023, all participating countries have reached an agreement.
Copyright
Ensuring the legality and compliance of the content delivered by neural networks is another rising challenge. The lawsuit from The New York Times is not the only one – George R. R. Martin and John Grisham were among the writers who sued OpenAI for the unauthorized use of their creations.
Initially, Valve was reluctant to publish games with AI-generated content in the Steam store, yet recently, it decided to lift the overall ban.
The debate is heated by court battles regarding the right of AI to be a creator. According to the US Supreme Court, for example, AI-generated work is not copyrightable. But the list doesn’t end here. What’s even more surprising is that there are AI-generated books published under the name of existing authors to hijack their fame and credibility.
AI Trend 3: AI impact on society
The widespread adoption of AI-powered solutions is already changing society. Workforce automation and augmentation are only top-of-the-head examples. According to a report issued by Bain & Company, up to 41% of processes in professional services, 40% in the media industry, and 39% in agriculture can be automated using generative AI. There were already examples of companies replacing up to 90% of their staff with AI-powered tools.
Also, there is a serious concern about generative AI’s impact on elections and politics. According to data gathered by Forbes, over 75% of consumers are concerned about misinformation spread by AI-aided groups.
Companies are aware of the problem and are shaping their policies to mitigate the risk. OpenAI has recently issued a statement about how the company is going to approach the elections in 2024 – these will include the US presidential elections in November, India’s general elections in May, and the European Union parliament elections in June, among others. Banning the Dean Phillips bot from the GPT store was one of the first (but undoubtedly not the only) election and AI-related incidents to be addressed.
AI Trend 4: Hyper automation and AI integration everywhere
Artificial intelligence, apart from changing the world, is also a business buzzword – albeit not a shallow one. This situation has made implementing AI and ML components in tools a good way to make a brand “fresh and tech-savvy.” This has resulted in introducing AI capabilities to tools that seem a strange pick.
Todoist, Notion, and Slack are tools that have implemented AI capabilities into their workflows. Microsoft is heavily pushing its AI copilots into its software, to the point of replacing the Windows key with an AI key on the keyboard.
The trend will also be reinforced by the multimodality of new models. AI tools are spreading out of their niches and starting to work effectively with multiple types of data, enabling them to support work in many ways – for example, aiding a writer with a text and inserting an infographic that is based on data cited in the text.
AI Trend 5: Environmental impact and chip shortages
Both businesses and personal users have shown enthusiasm toward these new tools. Yet AI comes with a not-so-hidden cost of the resources used to train it and later – to run it. According to Scientific American, data centers powering AI (among others) account for up to 1.5 percent of global electricity consumption.
The boom for AI-focused hardware results in record-breaking demand, with Nvidia shipping 1.5 million AI units per year. If these all run at their full capacity, they will consume 85.4 terawatt-hours yearly. This is more than Belgium consumed in 2022.
The AI community is addressing the issue. There are works to optimize the models to work efficiently on commodity hardware. The challenge is basically in reducing the power and memory intake to achieve comparable performance.
Also, leading model suppliers are working not only on their major developments but also on smaller models aimed at tackling narrow challenges. Phi-2 is a good example of a smaller, performance-optimized model.
Summary
We have the privilege of living in a time of great change. The technologies of the future, including talking computers and personal assistants powered by digital minds, are being introduced into our daily lives step by step.
The trends listed above are a reliable guide not to be lost or confused in this dynamic world of Artificial Intelligence’s current trends. If you have any questions or concerns regarding the role of AI in the near future, or maybe you just want to chat a bit, don’t hesitate to contact us now!
