AI Insights: Key Trends From the First Half of 2024

The year 2024 has been action-packed with AI launches and announcements. Just a few AI-related use cases from 2024 to illustrate how stunning AI can be:

• Runway introduced Gen-3 Alpha, a text-to-video AI generator capable of creating impressive, movie-like viral outputs with character consistency, camera motions, and scene transitions.
• Figure AI demonstrated its Figure 02, a next-generation AI-powered humanoid robot, featuring speech-to-speech for life-like conversations with humans and six RGB cameras with 360-degree vision to help it navigate the real world.
• Agibot, a China-based robotics startup, unveiled five advanced humanoid robots, directly challenging Tesla’s Optimus bot.
• Researchers have developed an AI model that can predict heart disease, diabetes, and cancer. The model is significantly outperforming existing methods.
• Verizon deployed AI and machine learning to predict and prevent fiber cuts, enhancing network protection efforts.

AI is now all over the news, and it is quite hard to keep up with the avalanche of new tools, facts, and figures. To grasp how the industry is changing today, we’ve pinpointed key AI trends and insights. Keep reading to learn more about them.

Trend #1 Tech Giants Rush to Deliver More Robust AI Models

Known for its successful large language models like GPT-4, OpenAI continues to dominate the AI landscape with substantial funding and widespread adoption across various applications, including chatbots and automated content creation​.

Meanwhile, tech giants like Google, Apple, Meta, Microsoft, IBM, and Amazon are heavily investing in AI research and development (R&D), increasing the AI market cap.

Here are a few recent AI milestones:

In December 2023, Google Deepmind launched the Gemini language AI model, making a step forward thanks to its multimodality (we’ll talk about this later) and contesting OpenAI’s dominance in the market.

In April 2024, Meta introduced its Meta AI assistant built with Llama3. Its creators claimed the model was smarter and faster than its previous series, surpassing many competitors in size and training data volume. Meanwhile, it preserved the open-source foundation, unlike Gemini or Mistral proprietary models.

In May 2024, GPT-4o was released in addition to GPT-4 (launched in March 2023). According to OpenAI, GPT4o is also multimodal (o stands for omni), yet twice as fast as its predecessor.

Mid-year, the experimental Gemini 1.5 Pro claimed the top spot on the AI Chatbot Arena leaderboard, surpassing OpenAI's GPT-4o and Anthropic's Claude-3.5.

Also, this summer, OpenAI set foot in the search market with its new SearchGPT.

Overall, the list of new AI solutions may be endless, as the quality of new networks changes from month to month, and more and more companies develop AI solutions for different needs.

Trend #2 AI’s Multimodality to Complete More Complex Tasks

AI systems are evolving from text-based to multimodal models, which means they integrate various types of information, including text, audio, and images. The new AI models move freely between natural language and computer vision tasks, allowing them to perform image-to-text, text-to-speech conversions, and more.

Multimodality enhances the contextual relevance of AI responses. It creates new opportunities for complex use cases where AI can handle versatile data input. For example, in areas like customer service, AI can analyze spoken requests, interpret documents, and evaluate facial expressions during real-time video consultations​​.

The major AI chatbots, such as Google’s Gemini (formerly Bard) and OpenAI’s ChatGPT, have multimodal capabilities. For instance, ChatGPT can describe images and answer questions related to an image description.

Another example is Microsoft’s Phi-3 family of small language models (SLMs), available on the Azure platform. The new Phi-3-vision model, introduced in mid-2024, can process both text and images to generate text responses.​ Phi-3 models are claimed to be powerful, cost-effective, and optimized for personal devices.

Trend #3 Smaller Yet More Efficient Language Models

The industry is shifting towards optimization for smaller, more efficient language models that can be deployed and managed more efficiently. Techniques such as Low-Rank Adaptation (LoRA) and quantization are becoming popular, enabling faster and more cost-effective fine-tuning and inference.

For example, in mid-year, Mistral Large 2 was released. This new AI model is claimed to be equal to, or even surpassing, the performance of the latest versions from OpenAI and Meta, despite having essentially fewer parameters. At just one-third the size of Llama 3.1 (405b with benchmarks comparable to GPT-4), Mistral Large 2 is claimed to be significantly more capable in code generation, mathematics, and reasoning.

On July 18, 2024, OpenAI released GPT-4o mini, a cost-efficient, smaller model in its GPT series that should replace GPT-3.5. GPT-4o mini outperforms GPT-3.5, while costing less than the company's other options. The new model is positioned as a competitor to other small language models, like Claude Haiku.

Trend #4 Democratization of AI

All the trends above — multimodality, shrinking resource requirements, and the open-source nature of many AI models — are crucial for democratizing AI capabilities. Democratization means that AI becomes more widely-available to the public, startups, and smaller players, reducing the dominance of large tech firms.​

Additionally, the use of APIs and microservices architecture is becoming prevalent. It allows tech companies to create complex custom AI-driven applications without substantial infrastructure investments.

These newer technological solutions ensure cost-efficiency. For example, for developers using OpenAI's API, GPT-4o is more cost-effective. As of summer 2024, it was available at a rate of $5 per million input tokens and $15 per million output tokens, while GPT-4 cost $30 per million input tokens and $60 per million output tokens.

Trend #5 GenAI Adoption Rate Is Growing 

The omnipresent AI is already making an impact on all spheres of business – and it is going to make much more.

According to surveys by McKinsey & Company, in 2024, generative AI adoption has seen a substantial spike, compared to relatively even adoption rates throughout several successive years. Around 72% of organizations are now using Generative AI in at least one business function, up by nearly 20% from the previous year. Significant increases were noticed across all regions, especially in Asia-Pacific and China.

The most common spheres of AI adoption in business are marketing and sales (34% of the respondents) and product/service development (23% of the respondents). However, AI can already cover practically any business function.

Even innovation groups use generative AI to enhance creativity. For example, incorporating image generators and text-generation tools in product development can transform traditional innovation workflows, helping gain insights and add user-friendly interfaces.

A notable trend is how AI content generation is gaining traction, with creative professionals voicing fears that AI will soon take over their jobs. Currently, more than 10% of search results are AI-generated, which means that the ranking URLs satisfy the searcher's intent and pass search engines’ quality requirements. Besides, enterprises deploy AI to scale content generation for personalized sales and other things.

Trend #6 Automation of Workflows with Chatbots and RPAs

The first and foremost application of generative AI is ensuring personalized experiences throughout the customer lifecycle. Smart AI-driven chatbots can guide users toward purchases, collect data, validate orders, and more.

Large E-commerce platforms, including Adobe Commerce, Amazon, and Shopify, add AI features to provide personalized customer service, assist with sales, show personalized search results, personalize content dynamically, and so much more.

The integration of AI-powered Robotic Process Automation systems (RPAs) allows companies to optimize workflows, reduce workload, eliminate human errors, and spare human workers for other tasks, such as overseeing the automated flows.

Robotic systems are becoming more advanced and autonomous, capable of interacting more effectively with the real world. For example, robots can now ask questions, autonomously master games like Minecraft, go online shopping, or assist with research​​.

Automation and robotization are surely going to transform workspaces. While it may make many workers redundant, it also creates new AI-related professions. To name a few, there is an increasing demand for AI prompt engineers, digital twin engineers, AI operations specialists, and others.

Trend #7 Growing Demand for Responsible AI

Despite so much techno-optimism, AI developers face increased disapproval and backlash. Critics blame GenAI for inaccuracies, or “hallucinations” in outputs, and accuse AI of being biased, or even outwardly overtly racist. Besides, there are fears that companies’ data may leak through employees’ use of AI. In addition, owners of intellectual property worry that LLMs may reproduce copyrighted material.

The point is that society and governments are not quite ready to cope with the numerous challenges posed by thihs fast-evolving technology. The number of AI misuse incidents has grown twenty-fold over the past 10 years, and as LLMs become more advanced, the more blatant these cases are. For example, this year, deepfake pictures of popular singers and politicians have gone viral on social media.

The issue has been aggravated by Sam Altman’s back-and-forths with OpenAI, his battle with Elon Musk, and the exodus of several top employees from the company. These and other events make people concerned about the future of OpenAI, now one of the biggest AI players.

There is a growing emphasis on the need to develop AI responsibly. Users need AI systems that address privacy, transparency, security, and fairness issues. Some global efforts are now being made to standardize safety protocols and ethical guidelines. For example, initiatives like the EU Artificial Intelligence Act aim to regulate the use of AI.​

Trend #8 Customization of Out-of-the-box AI Solutions

Being aware of the risks associated with AI, companies typically expect positive outcomes from integrating AI into their business operations.

Many organizations are leveraging off-the-shelf models, but there's also a significant trend toward customizing these models to meet specific business needs. According to McKinsey research, sectors like energy, technology, and telecommunications increasingly tailor models, or develop proprietary ones, to address unique challenges​.

This is especially relevant for legal, healthcare, or finance domains where specialized vocabulary and concepts may not have been learned by foundation models in pre-training. Meanwhile, LLMs for these areas can be small enough to be run locally. This approach enhances the relevance and effectiveness of AI solutions.

Bottomline

Generative artificial intelligence has become the fastest-developing industry ever. The global AI market size was estimated at $196.63 billion in 2023 and is projected to grow further.

These trends shaping the industry in 2024 are likely to stay for a while. Companies are competing to establish a firm ground with their products, and the very mention of AI can drive substantial investments in start-ups. The first half of 2024 has been marked by a sharp increase in generative AI adoption. Technological advancements have accelerated the arrival of more cost-efficient and better-performing AI models.

Have we missed any interesting AI trends? What news about AI have you found the most exciting? Share your ideas with us in the comments. Follow us to receive more reviews from our expert teams.