Remember the busy year 2023? Get ready because we’re jumping into the new AI wave, which will be even more significant in 2024. We’ll be figuring out new ways to use it and will have to tackle new problems. Let’s unpack the Generative AI trends you’ll want to keep an eye on!
A year after its introduction, ChatGPT has become a household name and the fastest-growing application ever. Midjourney has inspired our collective creativity and showed revenue reaching hundreds of millions of dollars. Character AI, an AI companionship social app, boasts record DAU/MAU metrics and deep engagement rates, reaching 2 hours per session.
According to the McKinsey 2023 global survey, 79% of people said they had some experience with Generative AI. For 22%, it has become a part of their professional toolkit.
Organisations, too, are now incorporating Gen AI:
- A solid 60% report integrating Generative AI into their operations;
- 40% of that group have also observed their companies boosting their AI budgets;
- And for 28%, the use of Generative AI is already on their board’s agenda.
So, as 2024 approaches, let’s prepare to meet the future of AI head-on. GenAI is not just here to stay but to lead the way.
1. Digital employees to work for us
2. Large models will get smarter, but fewer
3. Facing the challenge of chip shortages
4. Small and use case-specific language models will continue to bloom
5. Multimodality is the new black
6. RAG is what the enterprise needs
7. The most impacted industries will crystallise
8. Set up now for AI-augmented development
9. The rising AI risks will need a proactive approach
10. AI legislation and regulations will come into force
In the future, imagine starting your workday with a digital assistant powered by Generative AI. This assistant will be in your favourite communication apps like Slack or Teams and will manage your schedule and tasks easily. It will remind you to schedule a meeting, draft a follow-up letter afterwards, or create a list of arrangements for a new project.
Large Language Models (LLMs) can understand and generate natural language and execute tasks. However, they struggle to communicate with the outside world, significantly limiting their capabilities.
So, how can we make LLMs go from just answering questions to actively handling tasks?
The latest update from OpenAI provides the next step — the JSON mode for GPT-4 Turbo, which introduces the ability to return structured data and execute functions. This innovation is the first step towards integrating LLMs with the external world. Furthermore, tech leaders like Anthropic and OpenAI are developing models that can invoke external functions, allowing AI to execute tasks in the real world.
From a medium- to long-term perspective, Generative AI will be able to act like a personal assistant, enhancing human intellect within the workplace. Different technologies—systems, apps, and devices—will interact smoothly to boost productivity and innovation.
Few companies globally will possess the technological prowess and computing power required to develop frontier LLMs. Among notable entities are OpenAI, Google, Anthropic, Inflection, X, Meta, and several distinguished Chinese companies.
The progress is hugely determined by the number of Graphics Processing Units (GPUs) at a company’s disposal, as they play a crucial role in training large AI models.
This trend raises the stakes for open-source models, which risk falling behind in the race. Furthermore, it may cause segregation between foundation model providers specialising in scale and research and application layer companies focusing on product and UI development.
Why this trend is significant for businesses worldwide? While LLMs undeniably enhance operational efficiency, the most powerful ones remain accessible only through vendors’ cloud services.
Having to rely on third-party models, every company should now have a cloud LLM strategy.
Tovie AI’s GenAI product line can ensure your LLM operations are secure and your company’s sensitive data stays safe. As the industry changes, businesses must stay on top by adopting new and creative solutions that match the evolving Generative AI industry trends.
Contact us if you have questions or want to see how Tovie AI can enhance your LLM strategy.
And speaking of GPUs… As Forrester predicts, limited chip availability will temper AI expectations as processing capabilities hit a bottleneck in 2024. It may affect major tech players like Meta, OpenAI, Tesla, and cloud providers.
For example, the popular NVIDIA H100 AI processor is priced over $40,000 with demand-driven inflation. It also has a power consumption of 700 watts, making it both financially and ecologically abusive. As a result, in its annual report, Microsoft acknowledged for the first time the availability of GPUs as a potential risk factor for investors.
Companies will have to find ways to be more efficient to deal with the chip shortage. They’re considering using smaller AI models that don’t need as much computing power or trying out different ways of doing computations. Being adaptable and developing new ideas will be essential to keep AI moving forward in 2024.
4. Small and use case-specific language models will continue to bloom
Continuing from 2023, when we saw the rise of successful small language models designed for specific industries and tasks, this Generative AI trend is picking up speed.
Harvey, for instance, specialises in creating LLMs for elite law firms. Character AI and Ava focus on developing digital companions. Clinical Camel is an open-source medical language model that outperforms GPT-3.5 but is prone to hallucinations.
Smaller language models are popular because they are efficient, cost-effective, and customisable for particular jobs. Now, with the option to fine-tune models like GPT-3.5 and Llama-2, companies can adjust these models to fit their specific needs and enhance them based on user input. Platforms like Hugging Face, Cohere, and others make creating and using personalised small language models simple and affordable.
Exploring the business landscape in more detail, we find that AI chatbots are LLMs’ most popular use case. Powered by Gen AI and built on company-specific data, these chatbots offer a unique and tailored user experience without needing to develop an AI model.
Tovie AI introduces its Data Agent, a Gen AI tool that generates contextually accurate responses in a chat format. Integrated with various data channels and sources like Google Drive, Dropbox, SharePoint, CRMs, and MP3 audio files, Data Agent offers deep insights and accurate answers.
Contact us to learn more about Tovie AI’s Data Agent solution and how it can enhance your company’s Generative AI performance.
Most AI tools are specialised, focusing on tasks like text, images, or sounds. However, the latest trend is “multimodal” AI, where different types of AI are combined into various formats simultaneously.
This involves merging different models and creating interfaces that work in diverse ways, such as ChatGPT’s update that incorporates voice and image capabilities.
Other technologies, like Perplexity’s generative user interface, guide your search with interactive inputs, providing personalised answers. Similarly, the Inflection AI platform accurately recognises and analyses speech aspects like intonation, stress patterns, and vocal cues.
AI now doesn’t just understand your words but also sees the pictures you share and listens to what you say. This expanded capability opens the door to numerous possibilities, from creating content and interacting with smart virtual assistants to enhancing virtual reality experiences.
In short, multimodal AI will be one of the most promising trends in 2024, creating a more connected and exciting way to use the technology.
RAG (Retrieval-Augmented Generation) is one of the Generative AI trending tools. It enables AI models to consider information they weren’t trained initially on. Unlike fine-tuning, which modifies a model, RAG lets the model temporarily use external data for responses and then forget it.
Using external knowledge bases enhances answer accuracy and allows users to verify the sources of the AI’s responses. RAG diminishes the need for continuous retraining of models with new data, cutting computational costs.
RAG supports a more dynamic interaction than traditional scripted chatbots, which are limited and can’t adapt to new queries without manual updates. It allows LLMs to provide personalised replies by consulting the latest documents.
By grounding LLMs in external, verifiable facts, RAG minimises the risk of leaking sensitive data or ‘hallucinating’ incorrect or misleading information. For example, IBM uses RAG in its internal customer service chatbots, ensuring accurate, tailored responses by sourcing information directly from internal documents and company policies.
Despite challenges, especially with complex questions, RAG remains essential for mitigating the constant need for LLM updates for companies that have large databases.
According to the McKinsey report on the economic potential of Generative AI, about 75% of the value that GAI use cases could deliver falls across four areas:
- Customer operations
- Marketing and sales
- Software engineering
- R&D
The use case examples include supporting customer interactions, generating creative content for marketing and sales, and drafting computer code based on natural-language prompts.
Unsurprisingly, expectations for Gen AI’s impact are high across all sectors, with three-quarters of respondents anticipating significant or disruptive changes akin to industry competition in the next three years.
However, the level of impact is likely to vary, with knowledge-intensive industries that could see the most significant impact in their revenues are the following:
- High-tech
- Banking
- Life sciences (with pharmaceuticals and medical products in the first place)
For instance, in banking alone, the technology could potentially add $200 billion to $340 billion annually if fully implemented.
In contrast, manufacturing-based industries like aerospace, automotive, and advanced electronics may experience fewer disruptive effects. Unlike past technological waves that mainly influenced manufacturing, the shift is due to Gen AI’s expertise in language-based tasks rather than those involving physical labour.
Gartner has named AI-augmented development one of the top 10 strategic tech trends for 2024. AI technologies, such as Generative AI and machine learning (ML), are increasingly used to help software engineers create, test, and deliver applications.
According to Gartner’s predictions, by 2028, 75% of enterprise software engineers will use AI coding assistants, up from less than 10% in early 2023. Moreover, by 2025, half of leadership roles in software engineering will mandate oversight of Generative AI. This marks a paradigm shift in the skills required for future tech leadership.
AI-augmented development tools are becoming a trend because they allow software engineers to spend less time writing and debugging code. As a result, they can focus on more strategic activities, such as designing and composing impactful business applications.
These tools seamlessly integrate into developers’ existing workflows. They can generate application code, modernise legacy code, perform design-to-code transformations, and improve application testing.
The democratisation of AI has brought a new wave of challenges, urging businesses to prioritise clear risk mitigation strategies. Without proper safeguards, AI can generate compounding adverse effects that outweigh its positive performance.
The recent Risconnect report shows that most companies (93%) anticipate significant threats associated with Generative AI, signalling a universal awareness of this transformative but potentially unsafe technology.
Yet, only 9% of companies say they are prepared to manage Gen AI risks. The top concerns are data privacy and cyber issues (65%), employee decisions based on incorrect information (60%), and ethical risks related to employee misuse (55%). Copyright and intellectual property (34%) and discrimination risks (17%) also cause substantial worries.
Risconnect, the New Generation of Risk Report 2023
The gap between AI concerns and strategic readiness indicates that the technology is moving so fast that companies don’t know where to start. However, getting a handle on the risks of GenAI now is crucial to implement the technology successfully.
Addressing risks will be one of the leading Generative AI trends in the coming year. Adopting AI policies should first focus on protecting a company’s proprietary information, such as data, knowledge, and other intellectual property.
Using data masking technology helps to prevent data leaks with cloud-based AI systems. It hides your sensitive details, keeping them out of reach from external AI networks, and lets you control how your data is used. Thus, if sensitive information is detected, it is masked or blocks the processing of the request.
One of the advantages of data masking technology is cutting the costs of hosting and operating your own LLMs. But for enterprise settings that require even more strict policies, LLMs can be deployed on-premises in a private cloud. It guarantees companies complete control over how your data is handled, stored, and managed.
Tovie AI offers these and other solutions for the secure deployment of LLMs for businesses. Contact us to get a free consultation and discuss your organisation’s needs.
10. AI legislation and regulations will come into force
We will likely see new AI legislation in 2024. Following the signing of the Bletchley Declaration on AI safety by 28 countries, there is a clear global shift towards prioritising AI safety and regulation.
The UK government has outlined its objectives, and a pilot scheme is set to launch in 2024. Additionally, the Artificial Intelligence Act 2024 is set to come into force, indicating the UK’s commitment to regulating AI.
The EU policymakers have already agreed on the landmark rules within the AI Act that will impact businesses developing, selling, or using AI systems or components. Like the GDPR, it adopts a risk-based approach with fines reaching up to EUR 35 million or 7% of worldwide turnover.
The act will have a transition period of 18-24 months, giving companies time to act swiftly to comply with impact assessments, documentation, reporting strategies, governance programs, certification processes, vendor due diligence, and employee education.
Policymakers in the US are focused on regulating AI platforms to balance government use, algorithmic discrimination, and US global leadership in technology with national security priorities. Additionally, the 2024 presidential election is called the first “AI election,” so there is a specific focus on securing US elections from deepfakes and other AI-generated misinformation.
We are witnessing the evolution of the Generative AI market, moving from its initial buzz to tangible benefits and practical business applications.
Explore the current Generative AI map in this article, and let’s watch together how the field will adapt and evolve in 2024, keeping a keen eye on emerging Generative AI trends.
Now is the ideal time to start your Generative AI journey and reap its benefits for your company