/dq/media/media_files/2025/01/23/the-hotplate-of-2024.jpg)
It had AI, it had NPUs, it had new LMs, it had AI smartphones – and not just on the back-burners
A Flash-back is, often, the best way to get a Flash-forward. There’s no denying that the year 2024 was dominated by AI buzz for most of the time. But this year was also a crucial chapter that took forward AI in a really pronounced, and quite beyond the initial hype, way- in terms of scale, speed, spread, and specificity. So was the case with many other pages in the ever-expanding book of technology. Let’s read it through the lens of some hottest words from the shortlists of some pantheons of this year like the Oxford, Cambridge, Merriam-Webster and Macquarie Dictionaries. They, somehow, aptly describe some hottest technologies of the year while also giving a hint of what may come up ahead:
Yassification
Yassification means the act of giving something a glow-up or makeover – turning it more visually appealing. In that sense Gen AI went through a lot, yes really a lot, of yassification in 2024.
AI was probably the hottest topic in 2024, but very few people are talking about how AI will affect the data traffic in telecommunications networks and the need to increase investments in capacity in the future. - John Strand, CEO of Strand Consult
Interestingly, it was not just about text anymore. “Aside from text production, generative AI has also made outstanding progress in the fields of picture as well as video synthesis. For example, Midjourney and Stable Diffusion can produce extraordinarily realistic images from rather basic text commands. Additionally, such tools have evolved AI-assisted video editing which, as a result, provides ease in the production of high-quality videos for both individuals and corporations.” Weighs in Rethesh Nair, Senior Specialist Infrastructure Delivery, Publicis Sapient.
Nidhi Gupta, Senior Analyst, Technology Research & Advisory, Aranca notes how virtualisation, video encoding tools, and mini-language programs were made possible by the advancement of AI models like GPT-4 and DALL·E this year. “Notably, Generative AI started to aid with complex tasks like real-time language translation and AI-assisted medication development, promising to improve human skills. Despite ethical issues, the fast development of intrinsic AI and its broad use have strengthened its domination through good news, disinformation, and layoffs.”
It’s just not Gen AI but both predictive and generative AI, underlines Biswajeet Mahapatra, principal analyst at Forrester before calling it the hottest thing of 2024. ‘AI has a lot of potential and can change how software is built, tested, deployed, and maintained. It will also change how services and consulting are delivered. It promises to change human behaviours, societal norms, and government programs. It will lead to increased consumerisation, more affordable goods and services, higher productivity, and efficiency across the board.”
Enshittification
Welcome to the reality of platform decay, - now that is a pattern in which online products and services decline in quality. Social media and cybersecurity fall in this category – hands-down, literally. This year gave many instances of how fragile, tone-deaf and risk-prone safety has become – both in individual and enterprise digital spaces. Ransomware and AI-enabled frauds reared their heads with more chutzpah this year.
/dq/media/media_files/2025/01/23/AtXwdUJdwgI8Ter1WGqn.jpg)
In 2025, the rise of AI-powered cyberattacks, including Generative AI-enabled phishing and malware, is enabling larger, more sophisticated campaigns that evade detection and scale operations for widespread impact. - Sundar Balasubramanian, MD India and SAARC, Check Point Software Technologies
Sundar Balasubramanian, Managing Director for India and SAARC at Check Point Software Technologies rightly contends how AI continues to revolutionise industries such as healthcare and finance but also empowers cybercriminals to create advanced threats like deepfakes and vishing, which drive convincing social engineering scams. “Quantum computing promises unparalleled computational breakthroughs but poses significant risks to traditional encryption methods, necessitating a shift to quantum-resistant cryptography to secure sensitive data. Social media platforms, combined with Generative AI, are increasingly exploited for advanced impersonation and deepfake scams, using personal data to launch highly targeted attacks.”
Demure
Sober, grave, reserved or composed in demeanour- Coming all the way from the late 1400s. The word was in the buzz again since August 2024 after Jools Lebron used it in a viral video about make-up and wardrobe choices being mindful. So, what’s effortlessly cool and yet hot these days?
The most demure model on the runway of 2024 was, perhaps, none other than TPUs or Tensor Processing Units.
Generative AI undoubtedly stole the show. Its ability to create realistic and innovative content, from text to images and code, has revolutionized industries and captured the imagination of the public. - Parul Trivedi, Practice Director, Everest Group
TPUs – flanked by custom/proprietary AI chips- have taken over the baton from GPUs or Graphics Processing Units this year and have run straight into the Klieg Light of hardware. This year saw many significant rabbits being pulled out of many hats – from AMD (MI, Ryzen), ARM, Intel (Gaudi and Lunar Lake), Microsoft (Athena) AWS (Trainium and Inferentia), Tenstorrent and of course, Google (Axiom) and Nvidia (MGX designs and new Blackwell architecture). The latest being Google’s Trillium TPUs taking their aim at Nvidia’s Blackwell GPUs. If only the CUDA ecosystem and the strong software grip of Nvidia can be also easy enough to shake up! May be 2025 will show what’s up the sleeves of all these contenders and the UXL (Unified Acceleration Foundation) and Radeon Open Compute platform. Also, while Nvidia may clearly have imperialised AI silicon space in training workloads – there is a lot of room left on what’s coming next – yes, inferencing. It would be exciting to see how Groq and Cerebras surprise us on that front now. Who knows, NPUs (Neural Processing Units) would change this game again by the time we are ready to wrap up 2025!
Romantasy
A word that started circulating in 2008 when it was used as a genre label by Random House in Germany for the German translation of an English Romance Fantasy novel.
Chiplets happen to be the ‘Romantasy’ of the current silicon world. As Moore’s law hits new walls and slopes, as GPUs and TPUs elbow out CPUs and as AI workloads inflate- especially when inferencing changes the game of silicon in an unprecedented way, chiplets are turning into the pixies that take the industry into a new land. Chiplets remove the problem of cramming more transistors and complexity into yesteryear sizes and formats. Their modularity and ability to handle specific functions like data storage or processing signals and new kinds of workloads- make them the perfect answer for the post-AI world. Especially after the open-source standard Universal Chiplet Interconnect Express and moves made by AMD and Intel in this space. They are being explored for new use cases in robotics, space, automotive and AI as we fiddle with them with more courage and scale.
Manifest
The use of methods like visualisation and affirmation that help a person imagine achieving something specific and believe in its likelihood of happening. That’s what it is. The word jumped from use in the self-help community and on social media to being widely wielded across mainstream media and beyond, as celebrities such as singer Dua Lipa, Olympic sprinter Gabby Thomas and England striker Ollie Watkins spoke of manifesting their success in 2024. The oldest sense – which Geoffrey Chaucer ushered in the 14th century – is the adjective meaning ‘easily noticed or obvious’.
Guess how are humans manifesting ahead? As their digital twins. Yes, as holograms, AI replicas, virtual clones and near-close avatars. According to a new paper from a team including researchers from Stanford and Google DeepMind – the world is about to change a lot ahead with these undercurrents.
The acceptance of Bitcoin as reserves by many countries highlighted growing global confidence. - Himanshu Maradiya, Chairman and Founder of CIFDAQ
Agentic AI can be the next frontier of where AI goes next, after marrying many adjacent technologies. In the field of research, they are being called simulation agents, where they can be handy to conduct studies that would be expensive, impractical, or unethical to do with real human subjects. These are many levels up the vanilla AI we have seen so far- they are built to do things for humans, and not just converse with them. Announcements in this vein have been coming from Salesforce in the form of tool-based agents and also from Anthropic and OpenAI.
As we look ahead to 2025, the scope of these agents will expand dramatically, with multi-agent systems addressing complex challenges, such as simulating product launches or optimising marketing campaigns, augurs Arun Parameswaran, Managing Director - Sales, Salesforce India. We will witness AI agents working collaboratively, transforming productivity and redefining problem-solving at an unprecedented scale.
We may not be very far away from watching human personas created by studying individuals and used to mark a ‘proxy’ in virtual meetings or fight on a battlefield. Reportedly, companies as diverse as Microsoft, elevator maker ThyssenKrupp and Bosch have started exploring AR platforms and digital overlays between offsite workers and office staff. There is also the rise of Machine Customers - or non-human actors built using AI development. Consider a smart home appliance or an autonomous car- and it’s not hard to imagine someone else making decisions like shopping or entertainment or driving on your behalf.
Cocktail party problem
This word gained spotlight for explaining the very relatable difficulty of focusing on one voice when there are multiple speakers in the room. Nothing could explain AI’s struggles better. Well may be, another word – Slop. It was where originally overly sentimental literature was parked and later the spot moved to mean ‘any kind of nonsense’. In 2024, this feeling rose sharply in context to LLMs (Large Language Models)- the material used by such models because it is too low-quality and inaccurate and, hence, slop.
The rapid evolution of LLMs like GPT-4, as noted by Parul Trivedi, Practice Director, Everest Group, has enabled applications ranging from content generation and code writing to medical research and scientific discovery.
But LLMs have been running into some or other form of slop and noise after their initial dazzle. Bias, inaccuracy, discrimination, not enough data, not enough right data, and lack of context are becoming heavy problems to reckon with. This explains why 2024 saw a rise of nascent but formidable alternatives to LLM, such as SLMs or Small Language Models and Voice LMs.
IYKYK
If you know you know.
Here, it would be interesting to note all the usual roller-coasters that the world of crypto went through.
“The cryptocurrency market saw substantial growth in 2024, driven by Bitcoin’s 120 per cent rise and Ether’s 55 per cent increase. Key catalysts included the approval of Bitcoin and Ether ETFs, increased institutional interest, and a Bitcoin-halving event.” Captures Himanshu Maradiya, Chairman and Founder of CIFDAQ. He adds how Altcoins also gained traction, fuelled by renewed interest in DeFi protocols on platforms like Solana and Ethereum. “Institutional adoption grew via derivatives trading and innovations like liquid staking and restaking protocols. Additionally, the acceptance of Bitcoin as reserves by many countries highlighted growing global confidence. These factors indicate crypto’s maturation and integration into mainstream finance.”
Speedrun
This is a word used in the gaming universe, and it explains the ultra-fast effort required to complete a computer game or part of a computer game. So, it translates to any move that aims to complete something much faster than it is usually done.
Something that device makers did a lot this year with AI-Smartphones- trying to pull off a Parkour leap from yesteryear mobile phones and PCs to AI-devices. CoPilot+ from Microsoft, Apple’s Intelligence and Samsun’s Galaxy AI – with adjacent strides made by processors too (Qualcomm’s Snapdragon mobile updates) prove the pattern very clearly. In fact, a Canalys forecast tells we could see AI-capable smartphones grabbing 54 per cent market by 2028. What we need as the next Speedrun in 2025 is fraud-detection, deep-fake alert features built in AI smartphones too- as what Honor has teased with Magic Pro.
Wass Up Next
As John Strand, CEO, Strand Consult sets the tone right – “I hope that in 2025, people will start to look a little more critically at the hype that often exists around new technologies. Ultimately, it is a matter of translating new technology into revenue or reduced costs.”
Will that happen? Let’s see how ‘bussing’ 2025 gets. And how ‘Rizz’ AI stays. Till then, it’s ‘Toodles’ from all of us!
By Pratima H
pratimah@cybermedia.co.in