What I learnt about AI at Collision Conference 2023: From panel discussions to personal exploration and an ongoing AI learning journey

As you might have seen on my profile last June, I was fortunate to attend this year’s Collision Conf in Toronto, alongside Jennifer Kim , Partner and Executive Strategy Director at 21CB who took part in a panel about How brands can help design for real impact both online and offline with Tim van der Wiel.?

Collision Conference and AI?

The Collision Conference is one of the largest tech and entrepreneurship events in the world, which is organized annually and typically brings together thousands of attendees, including startups, investors, tech industry professionals, and thought leaders from various sectors. Currently based in Toronto, it has grown to become one of North America’s “most important technology marketplaces”.

It’s also worthwhile to note that Collision is part of the “Web Summit Group” which also organizes other tech conferences around the world, notably the Web Summit Lisbon (flagship event, November) which you might have heard of before. The event covers more than 30 topics, divided among 20+ tracks exploring every industry affected by tech.

Given my role as a Brand Strategist and the reality that you can’t follow everything, everywhere all at once, I focused my attention on the following tracks:

• PandaConf – The Marketing Track, exploring how technology has transformed the industry. Global giants, CMOs, international brands, and ad tech startups dive into the new landscape, discussing social media & influencer marketing, new frontiers in AdTech as well as purpose-led branding and marketing. (For obvious reasons)
• Venture, which gathers top investors at leading international funds, angels, accelerators and LPs shaping tomorrow’s investment trends, discussing the changing VC landscape, the venture playbook and emerging industries. (Since I work closely with tech companies given my proximity to Silicon Valley)
• And obviously the Center Stage, the main stages of the event, which serves as the focal point for high-profile keynote presentations, fireside chats, and panel discussions featuring prominent speakers and thought leaders from various industries.

AI Basics

It should be of no surprise to you, that this year’s buzzword is made of these two letters: AI. The advantage of hearing experts speak about the topic as opposed to sifting through my LinkedIn/Threads/X feed, meant that I got a more comprehensive perspective on the topic, straight seasoned engineers and coders, tech specialists and VCs with multiple AI companies in their investment portfolios.?

• As the more AI-versed people might have known already (I didn’t), Toronto is actually a big AI hub, as it’s the home of the renowned Vector Institute for Artificial Intelligence, which was co-founded by Geoffrey Hinton (you can listen to his whole take on AI here) and has received significant investment from the governments of Canada and the province of Ontario, as well as big tech companies like Google, Samsung, Uber and many more.?
• GenAI has been in the works for years and had a first breakthrough moment when one of the seminal papers related to AI and LRM (Long-Range Memory) – "Attention Is All You Need", commonly known as the Transformer paper ?– was released back in 2017 – and which is the foundation of the current LLMs such as ChatGPT.
• Most expert voices agreed that we’re still at the very, very, very early age of AI and have only seen the tippy tip of the iceberg so far.?
• AI is considered a “deep tech", involving complex scientific and engineering principles, cutting-edge research, and advanced development processes. There are probably less than 100 people worldwide who can actually properly build it, given it requires a high level of expertise, specialized knowledge, and significant resources to be developed successfully.
• While, at present, most AI systems are considered "Narrow AI" or "Weak AI", designed for specific tasks, most researchers and theorists goal is to reach "Artificial General Intelligence" (AGI), also known as "Strong AI" or "Full AI," which refers to a type of artificial intelligence that possesses the ability to understand, learn, and apply knowledge across a wide range of tasks. (Think DeepMind’s Flamingo capabilities which can be extended to multiple use cases)?
• Many still use AI as a toy (the mainstream and playful use cases of GenAI you might have seen so far), but it should be seen as a proper tool, given the wide range of application cases there already are:?- from automating the job in itself (eg. completing calls with customers),?- augmenting workflows (eg. gathering and extracting insights from customer feedback for developers),?- over optimizing infrastructure and supply chain challenges (eg. finding ways to cool down data centers)?- to accelerating medical research and breakthroughs (eg. DeepMind’s AlphaFold which performs predictions of protein structure within minutes, while phd research would have taken years)?- and many more.?
• The potential impact of AI is huge and could impact countries’ GDP, which has been studied and outlined in GitHub’s report Sea Change in Software Development: Economic and Productivity Analysis of the AI-Powered Developer Lifecycle, co-written by Dohmke and Marco Iansiti and Greg Richards from Keystone.AI
• While San Francisco and Silicon Valley still keep the title of the hottest AI places, it would be too reductive to only look there. Universities are usually the ones building a more scientific outlook on the matter and then find ways to monetize the findings. And while the West might look at larger innovations, dynamic markets like India and LATAM will seek to address and disrupt locally-relevant issues, i.e. redefining financial services.
• While the tech (and software) world was usually driven by replacement cycles of approximately 10 years, the rapid shipment of newer and self-improving versions of GenAI creates a much faster turnaround.?
• There are evidently many conversations about governmental regulations – understandably so – and the implications of Ethical vs. Unethical AI. Depending on the rulings, it could potentially create an innovation and competitive gap between the countries which will have to operate within stronger limits than others. While many people’s minds will go straight to Terminator scenarios, experts are still positive that AI won’t replace humans (yet) and that we shall remain in the driver seat – think of AI as your co-pilot or assistant.?
• Independently, big, AI-active companies are also looking inward to establish responsible ways of approaching and using AI, setting up review committees made of both internal and external machine learning, ethics and politics experts and? engineers.?
• I found it reassuring that the danger of biases and exclusion was top-of-mind for almost everyone, meaning that a significant effort has to be made to invest in education as well, to attract a more diverse range of people actually building the technology in order to properly represent the wider society.?
• Where will AI have the biggest impact? Ultimately, everywhere. For now it’s likely the developers which will use and benefit from it the most, but everyone agrees that everything will be impacted as the technology continues to evolve.?
• For business it means that AI is a threat and opportunity at the same time: the competition will be between businesses which embrace the technology, understand its opportunity, start using AI and find ways to deliver it to consumers – and those who don’t.?
• Differently from other historical tech cycles, being a second mover (eg. Apple before) might not be the best way to go, since AI gets better and better as it’s deployed. The faster you deploy, the faster it will be able to learn, evolve and grow.?

A VC specific lens on AI

I knew that one big difference between working in the US and Europe would be the direct impact of market dynamics and all things investment and VCs on the economy and our client portfolio. Being admittedly far from fluent in VC jargon, I couldn’t resist the opportunity to listen to top VC’s take on the current landscape and future outlook, especially on AI matters.?

• Investors believe that the whole AI movement is probably overhyped in the short-term, due to ‘mainstream’ use-cases of f.ex. ChatGPT, but will likely be underhyped in the long-term. It is yet to be seen if the current over-excitement will only be a healthy fuel or actually might stall the advancement?
• While a lot of money was raised in 2021, it hasn’t been deployed in 2022 – and these funds are now being poured into AI. This leads to quite high (and a bit unhealthy) AI valuations as many experts seemed to agree on.?
• A lot of the money is coming from more ‘generalist’ investors who are looking to fund diversify into tech (being uncomfortable with the performance of other industries), but who are at the risk of getting burnt, trying to approach the tech market the way they’d do with other verticals without properly understanding tech-specific dynamics and requirements.?
• While start-ups are inherently risky, AI will require even more rigor when it comes to the workflow of its different applications, given the necessity for stronger reliability of the product and data flywheel, as well as the infrastructure around it – and VCs are very carefully assessing every company’s value proposition and moat (i.e. in the tech VC context, a "moat" refers to a competitive advantage that a technology company possesses, making it difficult for other businesses to compete effectively against it.)
• There currently seems to be an extraordinary demand from customers for AI, which feels like there’s a great product-market fit with the potential of revenues growing exponentially, especially as there are more solid integrations and use cases being rolled out every day i.e. OpenAi’s ChatGPT API being used by big reputable companies.
• As mentioned in the section above, there are good chances that everyone will be using AI, but few will be able to provide it, given that 1) the talent pool is limited given the deep tech, 2) the time required to build it (just a reminder that it took years for OpenAI to build it as well), 3) computing challenges in the forms of global chip shortages and the need to access a lot of capital.?
• If you also feel like every company is advertised as being AI in some way today, you’re not alone. But investors are making a clear distinction between shallow applications of AI (building applications on top of existing products through an API i.e. “AI-powered”) and what I’m going to label “native AI” companies, which have developed their AI capabilities from the ground up on their own data and fully control its deployment.?

• As mentioned before, AI has been in the works for a long time, but what surprised VCs is how long it took the technology to mature enough to be released to the general public. We’re currently experiencing a powerful, but messy software which is still in its infancy and where we haven’t seen many proper implementations so far.??
• While we’re seeing seemingly incremental improvements so far, they could quickly turn into truly transformational changes - the question is when. The timeline is difficult to predict and we could be talking about either months or years.?

The more AI, the merrier?

Since admittedly hearing for the first time about the Transformer paper, LLMs, infrastructure and Narrow AI/AGI at Collision and understanding the true transformative potential of the technology, I decided to look deeper and further, beyond the almost playful ChatGPT and GenAI use cases that have taken over the internet and LinkedIn for weeks on end.?

I’ve signed up to pretty much every key GenAI tool there is, just to play around with it and get more familiar with the possibilities, limitations and differences. And acknowledging that it makes me not even close to an expert, I’ve decided to stay on it and to set the clear intention to read at least one AI-related article per day.?

Here are some other AI-adjacent resources and reads I’ve enjoyed lately:

• I’m currently reading “The Coming Wave: Technology, Power, and the Twenty-first Century's Greatest Dilemma” by Mustafa Suleyman (co-founder and CEO of Inflection AI , previously co-founded Google DeepMind ) where he establishes “the containment problem”—the task of maintaining control over powerful technologies—as the essential challenge of our age.
• The Financial Times published a brilliantly interactive article which explains how GenAI works thanks to the transformer
• Sequoia Capital shared an insightful take on AI’s $200B Question – and the need for the AI sector needs to find lucrative consumer-facing products and ventures to justify and offset the significant costs associated with running AI infrastructure.
• A research collaboration between Harvard Business School, Warwick Business School, and MIT found that BCG Consultants utilizing AI accomplished an average of 12.2% more tasks, completed these tasks 25.1% faster, and delivered results with a 40% increase in quality compared to their counterparts who didn't use AI. AI not only enhances performance but acts as a powerful up-skilling tool, particularly elevating those initially at lower skill levels, potentially bringing all workers to peak performance tiers. ?
• Google’s CEO Sundar Pichai's outlook on technology and AI for the company’s 25th anniversary. ?
• The Wall Street Journal reports that with artificial intelligence becoming the talking point in culture and technology, Senator Chuck Schumer (D-NY) convened a closed-door, all-Senate sit-down with the most influential stakeholders in the industry’s development and the leaders ensuring American workers aren’t left out of the equation. If talks are constructive, the meeting could supercharge an in-the-works bipartisan framework for AI laws that was recently announced. ??
• This article featuring physicist Max Tegmark (the scientist behind a landmark letter calling for a pause in developing powerful artificial intelligence systems) says competition too intense for tech executives to pause development to consider AI risks. ?

That’s it for now. Please do share any other interesting articles or resources you’d recommend I’d explore - always eager to add more to my neverending reading list.

-- I’d like to call out that the takeaways are, obviously, non-exhaustive but some key themes I heard during the panels and keynotes I was able to attend. Plus there might be chances that I misheard or understood some insights, feel free to point them out to me! --