From generative AI’s breakout year in 2023 to the rapid advancements in computer vision and natural language processing, this blog highlights the AI trends that matter—and will matter this year.
Our DevRel Manager, Harpreet Sahota, asked 12 industry experts four questions on what they think would shape the AI landscape in 2024, namely:
- Which AI trends are you keeping an eye on for 2024?
- How do you think these trends might reshape technology, industries, or our daily routines?
- Any challenges or big opportunities you foresee?
- What’s one aspect of AI for 2024 that has you excited or concerned?
Cut through the hype and stay ahead of the curve. Keep reading to learn what they shared.
Personalized AI, Yujian Tang
Which AI trends are you keeping an eye on for 2024?
“Personalization.”
How do you think these trends might reshape technology, industries, or our daily routines?
“Personalized AI is going to make our lives more convenient and do more thinking for us.”
Any challenges or big opportunities you foresee?
“The amount of compute required to run personalized AI may be overwhelming. Smaller, similarly performant models with caching is going to be a need.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“Bias in training data may become institutionalized.”
Smaller Multimodal Models and Broad-Purpose Robotic Systems, Charles Frye
Which AI trends are you keeping an eye on for 2024?
“The release of ever smaller permissively-licensed multimodal models with excellent reasoning capabilities, and a concomitant wave of tinkering, hacking, and discovery.”
How do you think these trends might reshape technology, industries, or our daily routines?
“In the short term, pushing the price of reasonable-quality perception and cognition down another order of magnitude or two, as we have seen in the past year, will unlock a number of use cases that are currently missing product-market fit due to price. For example, foundation models for AI in video games will look more like 70s arcade gaming once it hits a $1-$10/hr price point.
But more broadly: the last decade has seen tremendous advances not just in perception and cognition in the form of highly capable foundation models, but also in robotic bodies and their control, most prominently by Boston Dynamics.
I expect these two trends to merge and unlock broad-purpose robotic systems by the end of next year, in the form of a compelling, DIY personal robotics product, like a Roomba that follows natural language instructions or a mobile children’s toy/companion. This system may look more like the MITS Altair (too clumsy for anyone but enthusiasts) or the Xerox Alto (too expensive for mass marketing), but will serve the same inspirational purpose as those general-purpose personal computing platforms.”
Any challenges or big opportunities you foresee?
“As foundation model-powered systems penetrate the digital world further (posting to social media, engaging in transactions, etc.) and begin to tentatively extend tentacles into the physical world, the consequences of and debates around their impact will become sharper. Questions of agency and alignment — similar to but both more prosaic and more pressing than the same questions for theorized “”superintelligences”” — will be paramount.
Defining the terms and contours of this debate is a tremendous challenge/opportunity for our field. In my view, the more broadly the capabilities and benefits of these systems are distributed, the more public support they will enjoy, as with past technologies of great power and controversy, like cryptography and networking.”
Generative AI Agents, Vin Vashishta
Which AI trends are you keeping an eye on for 2024?
“Generative AI Agents that provide access to other models and apps and function as an expert assistant or advisor. Hyper-productivity tools (for example: enables one software developer to do the work of 2 or 3).”
How do you think these trends might reshape technology, industries, or our daily routines?
“Intelligent agents will change how people interact with technology and make expert knowledge more accessible. I see impacts on user interface design and customer expectations.
Hyper-productivity tools will reduce the number of technical resources necessary to deliver projects. Some companies will deliver more products faster and others will reduce headcount significantly.”
Any challenges or big opportunities you foresee?
“There are opportunities for Generative AI supporting simple use cases that weren’t feasible with prior technologies or ML approaches. The challenge is that most companies are ignoring what’s feasible today and trying to support use cases that Generative AI can’t. Basic implementations around search, application orchestration, and self-service tools will return significant value. Hype’s keeping business leaders and data teams from building in those directions.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“I’m excited about where research is headed for ensembles of smaller models that are orchestrated by LLMs. Each small model handles a niche domain or use case set with high reliability. The LLM breaks the user’s request into tasks and hands each one to the most capable smaller model or app. There are new product categories I’m excited to see delivered, but there are also a lot of applications in model self-improvement/self-retraining.”
Multimodal Embedding Spaces, Chris Brousseau
Which AI trends are you keeping an eye on for 2024?
“Continued multimodality, especially with embeddings like OneLLM and ImageBind. Also keeping an eye on multi-turn dialogue for agents.”
How do you think these trends might reshape technology, industries, or our daily routines?
“I don’t like technology that doesn’t directly make life easier for people. My hope is that these technologies allow for price decreases, overall deflation, and a much shorter work week for everyone.”
Any challenges or big opportunities you foresee?
“Monetization of LLMs, along with somewhat bumbling legislation, especially where data collection and processing is concerned are both huge challenges for this tech having a chance of benefiting people. I do see a huge opportunity for whichever companies become more open first. The less gate-kept knowledge there is in this area, the better, and the first company to open up will receive a huge market advantage.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“I’m truly so excited about multimodal embedding spaces. It makes me hopeful that we might bridge the gap between meaning and syntax in a real way by incorporating phonetics and maybe prototypical imagery.”
Generative AI, Susan Shu Chang
Which AI trends are you keeping an eye on for 2024?
“Generative AI.”
How do you think these trends might reshape technology, industries, or our daily routines?
“Simplify knowledge summarization and finding latent patterns in data. This helps the human side of things go faster (e.g. tools like Copilot).”
Any challenges or big opportunities you foresee?
“Domain knowledge will still be important to quality check and troubleshoot generative AI solutions.”
Multimodal and More Open Source Models, Merve
Which AI trends are you keeping an eye on for 2024?
“Multimodal models!”
How do you think these trends might reshape technology, industries, or our daily routines?
“Absolutely! It’s only the beginning, multimodality will change the way we all interact with everyday systems.”
Any challenges or big opportunities you foresee?
“The challenge is that the open-source models will catch up with closed-source models. Hopefully this will happen and maybe we could even see open-source models surpassing the closed source ones!”
What’s one aspect of AI for 2024 that has you excited or concerned?
“The research and open-source being harshly regulated to a point of losing competitive advantage and the few companies having a monopoly over AI is definitely a concern. However, the pace of the area is quite exciting!”
Huge and Cheaper Models, New Architectures, AI-First Products, and Voice, Peter Gostev
Which AI trends are you keeping an eye on for 2024?
“1) Huge models – the amount of compute available grew dramatically in 2023 (we just need to look at NVIDIA revenue), so I am very keen to see if we will get models with 10x or 50x more compute than GPT-4. It would also be a pivotal time to see on which path to AGI we are – if we’ll only get relatively mild improvements over GPT-4, then any form of super-intelligence will likely be a long way away.
2) Cheaper models – we already saw big reductions in prices – a year ago we had a barely usable GPT-3 model (Da-Vinci) that was the same price as GPT-4-Turbo now. I am hoping to see big reductions in prices for GPT-4 level models (maybe 50-80% in 2024), as well as 70-90% reductions in cost for GPT-3.5 level models – we are already seeing this with Mixtral 8x7b MoE model, where platforms are racing to cut inference costs to the bone.
3) New Architectures – I don’t expect them to make a difference in 2024, but it would be an important year for us to see whether new ideas (e.g. Mamba) would be actually better than existing transformer-based architectures. If they are, then problems like quadratic compute scaling for higher context will go away and we would see a big shift in certain aspects of LLMs (e.g. huge context lengths, cheaper training)
4) AI-first products – I want us to start moving away from chatbots and towards completely new, AI-first ideas. I am perhaps being optimistic in calling it a 2024 trend, as there are few and far between now (perhaps tldraw), but I am hoping that new creative teams will come together and start coming up with completely new ideas.
5) Voice – specific under-used category of AI-first products is voice. We have seen a little bit of that already with voice mode for ChatGPT, but I think there is a huge opportunity to create voice-based experiences for many. LLMs are brilliant at interpreting rumbling unstructured text and making sense of it – sounds like a perfect fit for humans who now won’t need to think too hard when they start engaging with products. I want to see new voice-based & LLM enabled onboarding experiences for new products – let’s stop getting our users complete our databases by hand for us during onboarding.”
How do you think these trends might reshape technology, industries, or our daily routines?
“In a way, the best kind of impact will be mostly invisible to users – things will work a bit better, quicker, smoother – if users will feel like they are speaking to AI or now need to do something unnatural to start engaging with an AI, that means we have some ways to go.
From the industry side, there is a question of how much is this innovation sustaining or disruptive. I suspect it would be largely sustaining for now – helping existing players become a bit more efficient and create nicer user experiences. But I’m really hoping for some big disruptions, maybe in law (Harvey.ai) or some other industry – we’ll need to see how it goes.”
Any challenges or big opportunities you foresee?
“Talent is a big challenge – despite all the hype around AI, we have a big shortage of people who know how to build AI applications and products. We need to get a lot more AI engineers with a good understanding of how these models work, what they are good at and how to tackle some of the issues.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“I am worried about some of the conversations around AI regulation – my view is that we are so early on with Generative AI, that regulating it in any meaningful way will likely be damaging. In Europe especially, we need to invest heavily to support talent & open source, we need to invest in compute and support start-ups – and I’m worried that the first instinct for us to regulate. Regulation is tricky, it is not clear at all that it even makes sense to regulate ‘AI’, we don’t regulate databases for example, we regulate industries where databases could be used in a harmful way – perhaps that’s the way we should go.”
Valuable Generative AI Production Use Cases, Rajiv Shah
Which AI trends are you keeping an eye on for 2024?
“Production use cases for Generative AI.”
How do you think these trends might reshape technology, industries, or our daily routines?
“If we don’t get enough valuable use cases into production, this is going to reduce interest in Generative AI.”
Any challenges or big opportunities you foresee?
“Identifying the next wave of generative AI use cases after chatbots, RAG, and code copilots.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“I am excited about it all – so much innovation going on.”
RAG and Smaller Specialist Models, Chris Alexiuk
Which AI trends are you keeping an eye on for 2024?
“RAG, Smaller Specialist Models.”
How do you think these trends might reshape technology, industries, or our daily routines?
“Swarms of LoRA fine-tuned models should likely outperform larger single generalized models – making these kinds of models more competitive at lower costs.”
Any challenges or big opportunities you foresee?
“Inference, inference latency, latency for agentic applications.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“The field seems to be mainly calming down – but it also feels like there’s some critical breakthroughs left and those will shake the industry.”
Wearables, AI-SecOps, DIY AI, and Continuous Intelligence, Serg Masís
Which AI trends are you keeping an eye on for 2024?
“1) XR+AI Wearables and HCI (Human-computer interfaces): Augmented Reality glasses, particularly Apple’s Vision Pro, are set to revolutionize digital interaction. Despite a lack of complete understanding of human neural activity, more advanced models trained on activated areas for most people and user-specific activity for intent will make it possible to automate many tasks with just the thought of performing it or generate artwork with a mental image of such artwork.
2) AI-SecOps and Proliferation of Zero Trust Architecture: In response to increased AI-driven (corporate) security threats, new cybersecurity professionals specializing in AI threats will emerge as part of newly formed AI-SecOps teams or existing SecOps teams. Given how susceptible humans are to GenAI trickery, zero trust architecture is expected to become more prevalent.
3) DIY AI: This is a bit self-serving because DIY AI is the title of the book I will release in the spring of 2024. But I genuinely believe that 2022-23 unleashed AI to aficionado types (many of whom transitioned from the crypto craze). However, 2024 will be the year that a broader group of early-adopter enthusiasts take AI into their own hands using ever more powerful consumer hardware to automate their own life through clever life-hacks.
4) Continuous Intelligence: Adopting more flexible data warehouse technologies and greater emphasis on observability will naturally lead to a more real-time augmented analytics view of business operations. It’s not a new thing, but changes in mindsets surrounding data will make executives more comfortable with changing the rigid breakdown of operations and corresponding metrics by month and quarter to a more flexible, nuanced, multi-layered, continuous view, especially when AI can find patterns that explain drops or increases in metrics on a granular level.”
How do you think these trends might reshape technology, industries, or our daily routines?
“1) Wearing AI: AI-powered AR glasses will not just be big for entertainment but also huge for healthcare and industrial applications augmenting human abilities to increase precision, speed and reliability of all kinds of tasks. Not to mention, folks that are disabled will benefit from a leveling playing field especially when coupled withHCI.
2) Weaponizing AI: New tools and methods will have to emerge as the cat and mouse game between organizations be it corporations or entire countries and their adversaries have to continuously one up each other. It will go beyond cyber security as IoT makes cyber physical systems and increasingly connected people more of a target.
3) Playing with AI: The same way people were playing with making their own websites back in the early 2000’s, people will now be making their own chatbots, and computer vision deep learning models. The least technically savvy will use cloud-based tools for this and the most savvy will learn how to do it from scratch.
4) Real-time Insights from AI: The deluge of information coming our way has become untenable for analysts to crunch at so many different levels and time scales. More organizations will adopt an approach that makes analyzing a real-time 24/7 endeavor and it will impact many business practices from accounting to supply chain to marketing.”
Any challenges or big opportunities you foresee?
“Challenges: CaaS (Cybercrime as a service) using Generative AI to sabotage and defraud businesses and individuals, swing elections and cause geopolitical upheaval. Physical limits of chip design, supply chain issues amid geopolitical problems and data center growing energy consumption will cause the machine learning inference to reach a limit.
Opportunities: first and foremost, to standardize, regulate and govern AI. From a corporate standpoint, this means embed Quality Assurance (QA for AI) and User Experience (UX for AI) in every aspect of the design and operation of AI systems but also certify that interoperability between systems ensures scalability, reliability and affordability. Society could demand that their data ownership/security and privacy is respected and algorithmic fairness is ensured. This would be a long-term win for trust and to reduce friction to accelerate adoption.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“I’m excited for the advantages the technology will bring to so many fields from entertainment to cutting-edge science while concerned that our society and institutions won’t keep up or use it for harm. As the quote by Edward O. Wilson goes “The real problem of humanity is the following: We have Paleolithic emotions, medieval institutions and godlike technology.”
Multimodal LLMs, MoE, and RAG, Shubham Saboo
Which AI trends are you keeping an eye on for 2024?
“1) Multimodal Large Language Models (LLMs): These models integrate multiple types of data inputs, such as text, images, and audio, to understand and generate content across various formats, enhancing AI’s ability to interact in more human-like ways.
2) Mixture of Experts (MoE): MoE models distribute tasks across multiple specialized sub-models (experts), each trained on different data subsets or tasks. This approach optimizes efficiency and performance, especially in handling complex, diverse, or large-scale AI applications.
3) Retrieval-Augmented Generation (RAG): RAG combines the strengths of pre-trained language models with external knowledge retrieval, enabling AI to generate more accurate and contextually relevant responses by accessing a vast range of external information sources.”
How do you think these trends might reshape technology, industries, or our daily routines?
“1) Enhanced Human-Computer Interaction: Multimodal LLMs and technologies like RAG and MoE models will make interactions with AI more intuitive and human-like, transforming user experiences in technology, from smarter virtual assistants to more interactive and responsive applications in education, entertainment, and customer service.
2) Data-Driven Decision Making: The integration of these advanced AI models in industries will lead to more informed and efficient decision-making. By synthesizing complex data sets, businesses can optimize operations, personalize marketing strategies, and innovate in product development.
3) Personalization and Accessibility: Daily routines will become increasingly tailored and efficient as these AI models enable more personalized content, recommendations, and services. This could range from customized learning experiences to AI-driven health monitoring, significantly enhancing the accessibility and relevance of digital services.”
Any challenges or big opportunities you foresee?
“Challenges
1. Ethical and Privacy Concerns: As AI becomes more integrated into daily life, issues surrounding data privacy and ethical use of AI will become more prominent. Ensuring user data is protected and AI is used ethically will be a significant challenge.
2. Complexity and Resource Requirements: The development and deployment of advanced AI models require significant computational resources and expertise. Balancing these demands with sustainability and accessibility, especially in less developed regions, presents a substantial challenge.
Opportunities
1. Innovation in Healthcare and Education: These AI technologies offer opportunities for groundbreaking advancements in personalized medicine and adaptive learning systems, potentially transforming healthcare and education sectors.
2. Business Optimization: AI can significantly enhance business efficiency, from automating routine tasks to providing deep insights into customer behavior and market trends, offering substantial opportunities for business growth and innovation.”
What’s one aspect of AI for 2024 that has you excited or concerned?
“One aspect of AI for 2024 that particularly excites me is the advancement in multimodal AI systems. These systems, which can process and understand multiple forms of data like text, images, and audio, promise to significantly enhance the way we interact with technology. Here’s why I find it exciting:
1) Human-Like Interaction: The ability of these systems to understand and generate content across different modalities makes interactions with AI more natural and intuitive, closely resembling human communication. This could revolutionize user interfaces and experiences across various platforms, making technology more accessible and user-friendly.
2) Creative and Educational Applications: Multimodal AI has immense potential in creative fields, like generating art or music, and in educational tools, offering interactive and engaging learning experiences. It can cater to different learning styles and aid in creating more inclusive and diverse educational content.
3) Cross-Disciplinary Innovation: This technology has the potential to spur innovation in numerous fields, from enhancing medical diagnostics through image and data analysis to transforming media and entertainment with interactive and personalized content.”
Small Powerful Models, Inference Cost Efficiency, Self Hosted Deployments, Yonatan Geifman
Which AI trends are you keeping an eye on for 2024?
As generative AI matures from experimental stages to full-scale production, a significant shift is expected towards smaller, more specialized models. This movement is driven by the need for more efficient and effective performance in specific applications.
What drives this shift? Firstly, smaller models, tailored for specific tasks, are predicted to deliver better performance. This is due to their ability to focus on a particular domain or function, leading to more accurate and relevant outputs. In addition, the adoption of smaller models is also motivated by their cost-effectiveness. Because these models are designed to be cost efficient, they require less computational power, reducing both operational expenses and the environmental impact associated with larger models.
How do you think these trends might reshape technology, industries, or our daily routines?
The increased availability of smaller and cost efficient models has the potential to allow more teams to build and scale advanced AI-powered applications and solutions, democratizing access to AI capabilities, fostering innovation, and enhancing efficiency across industries. Additionally, the more effective and accessible models become, the more they can support the development of more personalized and adaptive AI systems, and contribute to addressing a wider range of business or societal challenges.
Any challenges or big opportunities you foresee?
Today’s landscape of Large Language Models (LLMs) and other generative AI technologies demands substantial computational power, contributing significantly to the current scarcity of GPUs in the market. This computational intensity underscores the urgent need for innovative solutions that facilitate the operation of these models on a wider range of affordable hardware platforms. Such advancements would democratize access, enabling a broader spectrum of developers and AI teams to leverage, scale, and integrate AI capabilities into their projects seamlessly. Additionally, as the adoption of these AI technologies continues to expand, there’s a growing concern among companies about the potential risks these models pose to their data and intellectual property (IP). In response to this, businesses are increasingly seeking ways to exert greater control over their AI models, aiming to enhance data security and privacy measures. This shift reflects a broader recognition of the importance of safeguarding sensitive information while harnessing the benefits of AI innovations.
What’s one aspect of AI for 2024 that has you excited or concerned?
I’m very excited to see the growing innovation and new practical use cases that are being developed by companies in the space. Deci’s models and inference infrastructure is used by teams across industries including financial services, developer tools, Sales and marketing platforms to name just a few. We take immense pride in being the preferred technological partner for these teams as they embark on their journey to build Generative AI solutions that can be seamlessly integrated into everyday tasks on a cost-effective, scalable basis.
That’s a wrap!
With generative AI being one of the popular trends, Deci offers solutions to organizations that have plans of launching generative AI applications in 2024 and beyond. These include robust foundational LLMs and developer tools such as Infery, an inference SDK, that can accelerate deployment and enable cost-effective scaling.
Overall, we’re in for an exciting ride this year – full of advanced technologies that foster a collaborative relationship between humans and AI. But, while these breakthroughs offer unprecedented opportunities, challenges persist, necessitating ongoing efforts in balancing innovation with regulation, privacy protection, and bias elimination for responsible AI deployment. Like the previous years, the landscape in 2024 signifies another transformative era where AI acts as a powerful and responsible tool in shaping the future.
