2022 was a big year for the advancement of AI/ML with new innovations and more adoption than ever before. So as we move into the 2023 we asked leading AI experts to recap on last year and let us know what they thought was the biggest highlight for AI / Machine Learning. Ofcourse, Generative AI has undoubtedly been a hot topic as well as the viral emergence of ChatGPT by Open AI, but we also had other areas. Check out what the experts had to say!
For me, Generative AI centered around Diffusion models is a defining trend in 2022. Now we can generate a large diversity of even unseen/unrealistic images (e.g. Darth Vader Biking in a Park) with just textual cues.
- Aayush Prakash - Engineering Manager, Meta RealityLabs
Generative models have been hard to ignore. Besides direct consumer applications, it will be interesting to see how they may apply to other ML settings, such as generating cheap training data.
- Mathew Teoh - Senior Machine Learning Engineer, LinkedIn
Generative AI became more mainstream in 2022 with the image generative models of DALL.E and Stable Diffusion and more recent preview release of ChatGPT. Lensa AI app which is a portrait generating app based on Stable Diffusion reached top of iOS App store and sets tone f+or more products and services building on top of large scale generative models.
- Roopnath Grandhi - Product Leader, Entrepreneur, AI Leadership, Johnson & Johnson
Generative AI. The evergreen for me in the last decade has been the natural language processing, broadly construed. I am very passionate about it because it is the most exciting manifestation of human intelligence - this language capability is what differentiates us the most from other animals. This, of course, has been connected to the thought and is highly philosophical, but it also results in concrete products and, in particular, is at the very core of search.
- Richard Socher – CEO, You.com
Generative AI has been one of the biggest highlights of machine learning/deep learning in 2022. It refers to a set of techniques that enable AI algorithms to autonomously create synthetic data that is indistinguishable from real-world data such as images, text, audio, video and more. For example, it can be used to generate fake malicious content to test the robustness of security systems, create realistic virtual environments for gaming, and inspire artists to accelerate their creative work. Overall, generative AI has the potential to revolutionize many industries and bring about a new wave of innovation.
- Amey Dharwadker - Machine Learning Tech Lead, Meta
The advances in generative AI are amazing: human-level text, code and images. To be sure, AI is still work in progress, sometimes it generates silly text, incorrect code and bad images so we cannot use it where trust is required in the information it generates. But we all know tremendous progress has been made even if we cannot quite measure it objectively or fully trust it yet.
- Apostol Vassilev - Research Team Lead; AI & Cybersecurity Expert, National Institute of Standards and Technology (NIST)
ChatGPT. Like many others in the industry, I was obsessed with talking to ChatGPT to figure out its capabilities and consistently fascinated by the interesting interaction other people had with it. It’s clearly not close to AGI to myself but seems close enough to many people.
- Zhiyuan Zhang - Engineering Manager, ML Serving Platforms, Pinterest
It has definitely been the release of ChatGPT bot by OpenAI that crossed 1 million users in 5 days. It went viral and the users were not just tech enthusiasts but from all different kinds of professions who looked forward to giving it a try and sharing their experiences with the world. Not just enthusiasts but also skeptics rushed to provide their feedback and opinions on the technology that seems to be a big breakthrough in the space of Conversational AI. Even though OpenAI has not open sourced the technical details around it, it has pointed towards the fact that ChatGPT is a sibling model to InstructGPT, which is trained to follow an instruction in a prompt and provide a detailed response.
- Jyoti Mishra - Senior Data Scientist (NLP), Peakon, A Workday Company
The progress made by the open source research community to train and release large models that were previously only available in a few labs.
- Karl Willis - Senior Research Manager, Autodesk
In one word: transformers. While the theory has been around for a few years – the original transformers research paper came out five years ago – 2022 was a breakthrough year for consumer and commercial applications of transformers. Transformers are a type of deep-learning model focused on sequence-based data, like natural language. They originally were applied to problems like machine translation and text completion. But in 2022, the applications of transformers exploded. They emerged as a core part of groundbreaking text-to-image models, like Stable Diffusion, Midjourney, DALL-E 2, and coding completion tools, such as Codex, CodeWhisperer. But what’s even more exciting is that training transformer models with additional training data and learned parameters inside the model, or “nodes” inside the neural network, has led to the emergence of new, powerful, and unplanned capabilities. For example, language models have acquired the capability to accurately do arithmetic.
- Sam Stone - Director of Product Management, Pricing & Data, Opendoor
The biggest highlight for me is the continued investment in improvements in the technology. The language recognition is only improving and with the ability to choose different speech models based on business/customer need will only lead to better Customer engagement.
- Ross Parkes - Product Owner – Automation, HomeServe USA
The biggest highlight has been adoption by major corporations. This opens the door for investment into the conversational AI space and we'll see things we haven't seen before.
- Sonia Talati - Senior Manager - Conversation Design, GoDaddy
To learn more about these topics and the key trends, challenges and opportunities for the future you can join these experts at one of our upcoming 2023 events.
To celebrate the New Year we're giving you 30% off all passes to our 2023 events in our New Year Sale, just register your place with code NEWYEAR.
- Women in AI & Data Reception - 24 January, London, UK
- Virtual AI Summit – 30-31 January, Online
- AI Summit West – 15-16 February, San Francisco, USA
- AI in Finance Summit London – 25-26 April, London, UK
- AI in Finance Summit New York – 19-20 April, New York, USA
- Conversational AI Summit - 17-18 May, London, UK
- Deep Learning & CV Summit - 13-14 September, London, UK
- Berlin AI Summit - 11-12 October, Berlin, Germany
- AI in Healthcare Summit – 18-19 October, Boston, USA
- Montreal AI Summit – 1-2 November, Montreal, Canada