How artificial intelligence surprised the world and the editorial team in 2024. A year in review

2024 marked a significant milestone in artificial intelligence and neural network development.

Last year's predictions have come to fruition, and new trends and development directions for the next decade have emerged. Can you now imagine life without the GPT bot, always ready to answer any question you have? Or without AI artists who can create an image or a complete illustration in just seconds? Precisely!

As 2024 draws to a close, it's time to reflect on the major achievements in artificial intelligence that will leave a lasting impact for years to come and examine how the market has evolved. The Hitecher editorial team dives into this—read on to discover more.

Google's all-encompassing Gemini 2.0 model and the new generation of GenAI

We decided to start with the giants whose dominance in 2024 remains beyond dispute, particularly the world's largest company, Google. Their new AI model, Gemini 2.0, has reached a level of sophistication that allows it to replace humans in almost every area of life. Gemini 2.0 is genius: it can produce text, sound, and images and perform routine tasks involving information search and processing with its cutting-edge multimodal capabilities, which were unachievable just a year ago. Essentially, this AI can simultaneously connect to and work with multiple independent information sources, like a person sitting in front of a computer screen.

Like any company aiming to dominate the market, Google prioritises the commercial success of its products. For a product to be profitable, it must be in demand and useful. Throughout 2024, Google has been developing its AI infrastructure. Although Gemini 2.0 is still in the assessment phase, Google CEO Demis Hassabis already describes it as “comprehensive”.

Primarily, Google is banking on the “agent” function of its AI, where bots perform tasks on behalf of users and interact seamlessly with clients without revealing their automated nature. For instance, within the Gemini ecosystem, Project Astra can recognise and locate lost objects. Similarly, the Jules service aids developers in identifying and fixing 'broken' code. Google Project Mariner, an extension for the Chrome browser, can literally browse the internet for the user.

These agent AIs are expected to continue expanding and improving throughout 2025, becoming an integral part of our daily lives (or so we hope).

Models like GPT-4 and Claude 3.5, along with OpenAI o1, have reached new milestones in the realm of complex open tasks, even showcasing reasoning abilities. Their future applications could extend to fields like biology and neuroscience. However, in 2024, generative AI (GenAI) models, including the ones above, were primarily used in marketing and sales (66%), customer service (54%), research and development (49%), and IT (31%). Their primary roles in these areas were generating images, audio, and text, including educational content such as training courses and reports.

DeepMind in healthcare and scientific computing

Healthcare is a vital area for the development and use of artificial intelligence, with its potential widely recognised. Modern AI versions can diagnose and detect diseases from test data much faster than humans. Consider the story of a child in the USA last year who saw 17 doctors over three years without getting a correct diagnosis until the desperate mother consulted ChatGPT with a description of the symptoms. The bot provided the proper diagnosis, ultimately saving the child's life.

In 2024, AI made substantial strides with the DeepMind algorithm, which can predict three-dimensional protein structures, read information from RNA molecules, and perform numerous other medical analytical undertakings. Although DeepMind was introduced in 2017, it wasn't until 2024 that it began successfully predicting complications in diabetes patients, supporting the development of individual treatment plans, and examining biopsy and MRI data with accuracy and objectivity that surpasses human abilities. Furthermore, DeepMind can analyse and predict drug interactions and cell behaviour in countless scenarios and work with not only proteins but also amino acids.

Scientists from the University of California have introduced an AI capable of predicting the onset of Alzheimer's disease up to seven years before the first clinical symptoms begin. This AI relies on passive and secondary indicators such as cholesterol levels, hypertension, vitamin D deficiency, erectile dysfunction in men, and osteoporosis in women. Similarly, this AI can predict the onset of multiple sclerosis, a chronic autoimmune disease of the central nervous system, due to its varied symptoms, which are challenging for humans to track.

In 2024, DeepMind and its applications transformed mathematical science. For example, AlphaProof, which influences a pre-trained language model, can generate multiple solutions to any problem and independently verify them or even disprove them. AlphaGeometry 2 can solve complex problems involving the movement of objects and the relationships between distances and angles. Both algorithms successfully passed the International Mathematical Olympiad 2024, scoring 28 out of 42 points and solving four out of six problems, effectively earning silver medals alongside human participants.

The emergence of “safe products” and the need for AI accountability

The collective worth of companies engaged in AI development has now reached a staggering $9 trillion. With this tremendous growth comes mounting global concerns about the safety of neural networks and the priority for legislative regulation. A pressing issue is how to govern copyright laws, which artists have claimed are frequently violated by AI systems like Midjourney and DALL⋅E.

Consequently, over 127 countries have enacted legislation addressing artificial intelligence. For instance, a bill passed by the European Parliament aims to limit the use of facial recognition technology due to potential abuses that could infringe on privacy and confidentiality. This initiative also bans AI from gathering biometric data from social media and identifying reactions, even in criminal investigations, where law enforcement agencies have considered employing AI similarly to lie detectors.

According to the American technology company Gartner, businesses are increasingly concerned about controlling AI and regulating its scope of application. In this situation, 2024 could witness a shift towards open models like GPT-J, which are easier to configure, redirect, and verify. This shift could enhance opportunities for collaborative tasks and research.

What significant developments occurred in AI in 2024?

In summary, the Hitecher editorial team has also identified and actively approved the following achievements and transformations in the area of artificial intelligence and neural networks in 2024:

The increase in AI users. In 2024, the number reached a record 314 million, according to official and monitored data. Remarkably, every fourth person living in major cities interacted with AI, either through work or leisure. This signifies the widespread integration of AI not only in business but also in daily life.
The proliferation and mainstream acceptance of AI-generated content are on the rise. Forecasts suggest that by 2026, around 90% of internet content will be crafted using AI. In 2024 alone, the volume of this content saw a substantial uptick, even surpassing the levels observed in 2023.
Increased investment in the industry. In the second quarter of 2024 alone, investments in AI startups reached $24 billion, doubling the amount from the first quarter. The AI market's value is now estimated at approximately $300 billion, not including the worth of the companies themselves, as previously mentioned.

However, these are merely statistics. Simultaneously, in the realm of AI:

Competition is intensifying, and the features between different models are fading. The average user, for instance, may not notice the difference between ChatGPT and another bot based on OpenAI. Neural networks are so similar that they utilise nearly identical technologies, making it difficult to determine which is more advanced. This situation complicates the market in terms of advertising campaigns and ratings, posing potential future challenges for major companies like Google, which highly value their unique identity and leadership position (it’s no coincidence they have been clashing with Amazon for years).
Achieving “human-like” thinking is a key focus in the development of modern language models. Increasingly, scientists and startups strive to transform neural networks from mere mechanical tools into entities that think and react like living beings, a prospect that can be somewhat unsettling. To achieve this, multimodality is being actively developed and implemented: AI is no longer confined to language but is now demonstrating capabilities in mathematics, biology, and even genomics.
The cost of using AI is decreasing. Ironically, the development of powerful neural networks demands minimal computing power (after all, the purpose of AI is defeated if it is also highly energy-consuming). Consequently, the potential for AI to replace human resources is increasing. However, according to the Massachusetts Institute of Technology, only 23% of professions can currently be realistically replaced by AI, so there's no immediate cause for concern.

An increasing number of AI development startups are gaining recognition and substantial investments from large companies and investors. As a result, many new neural networks are emerging in the market. For 2024, we highlight and recommend that you try the following:

ChatGPT-4 — an enhanced language model that has proven to be an excellent learning assistant!
DALL⋅E 3 — the same image generator, but with enhanced detail and a much better grasp of instructions.
RunwayML — a tool for creative content that can, for instance, generate videos from text, remove backgrounds, animate objects, and more.
Jasper — a high-quality copywriter that maintains the originality of texts.
Synthesia — one of the top video generators with avatars this year, allowing you to save on filming and avoid the need to hire a professional presenter.
DeepL — AI for translation, supporting over 30 languages, with excellent contextual understanding.
Grammarly — reviews your texts for grammar, punctuation, and other errors.
Stable Diffusion — an open-source image generator.
EleutherAI — a research-focused scientific model capable of creating EAI models, solving maths problems, and much more.

In summary, 2024 has proven to be a groundbreaking year for numerous AI projects and, most importantly, for their communities—individuals who, like the Hitecher editorial team, believe that AI is the future. If you're still sceptical and think AI won't change our world, take another look at this article and read it again!