Big, big data is needed to power worthy AI predictive models, but that is still a work in progress, says a former lead investor in Google DeepMind. Pamela Whitby finds out more
Humayun Sheikh does not dispute the fact that Google DeepMind is a leading light in artificial intelligence research and its application for positive impact. As a former lead investor in the company, who exited when Google snapped it up in 2014 for a reported $500-million, that would be a bit ungrateful. After all, Sheikh had the opportunity of working alongside Demis Hassabis, the British artificial intelligence researcher, neuroscientist, video game designer, top gamer and entrepreneur, who co-founded what is today recognised as a world’s most innovative AI company.
What he does dispute, however, is whether new technologies like AI, machine learning and blockchain are truly able to deliver commercial value to small and large companies alike. And at the present, the answer is no.
“The only companies able to run effective AI algorithms and gain insights from their data are those with massive data stores and the opportunity to invest large sums in the right people and tools to analyse them,” says Sheikh.
In other words, the giant Internet companies, like Google, Facebook and Amazon!
He continues: “Building a neutral system that is open for any person or company to take part in is currently impossible, and the potential benefits of AI research is [still] closed off to many.”
Building a neutral system that is open for any person or company to take part in is currently impossible, and the potential benefits of AI research is closed off to many
Humayun Sheikh, co-founder and CEO Fetch AI, and former lead investor in Google DeepMind
That’s not to say, however, that big and small travel companies alike aren’t getting in on the AI and blockchain action. At upcoming EyeforTravel Amsterdam 2018, delegates can expect to hear from Wijnand de Groot, KLM’s VP of digital marketing and sales, about how the airline is using everything from personalised digital marketing to AI-supported social media to improve the customer journey. Meanwhile, Amer Mohammed, head of digital innovations at Stena Line will also share insights into their decision to appoint a dedicated Head of AI.
Arguably, even more out there is Air-France-KLM’s decision to appoint Iris Taguet as ‘head of blockchain programme’; she will also be speaking in Amsterdam about viable use cases for the much-hyped technology.
Clearly then, there is a lot of drive, and a lot of hype around what new technologies can deliver, and many travel brands are already investing.
However, Sheikh believes that some projects may not be as focused as they could be. The two questions, he argues, that project teams need to answer, are this:
- Is it bringing costs down?
- Is it removing friction?
“If those two things are not taken care of, then it’s not worth doing. This is pure economics,” he says.
The desire to commoditise AI and make it more accessible and affordable is what spurred Sheikh, along with former DeepMind head of software design Toby Simpson, and Thomas Hain, a Sheffield-based professor and scientist in advanced machine learning, to co-found Fetch AI – an acronym for ‘framework for economic transaction and communication hierarchy.
Backed by Outlier Investors, which invests at the ‘convergence of blockchains with AI, the Internet of Things and robotics’ and touted as one of its ‘most eagerly anticipated projects of 2018’, the 25-strong Cambridge-based company came out of stealth mode in March.
AI has the potential to double what is currently achievable using traditional analytic methods
While the startup is not focused exclusively on the travel industry, the Fetch team recognises that there are significant pain points to address, which new technologies can help to ease while also delivering cost benefits. A recent study by McKinsey found that AI, for example, has the potential to double what is currently achievable using traditional analytic methods, amounting to between seven and 12% of total revenue.
Unsurprisingly then, Fetch was keen to make a move in travel, and in August it announced a partnership with wi-Q Technologies, a cloud-based mobile ordering tech firm focused on hospitality. Together the two companies will deliver what they claim will be the world’s first ever autonomous economic agent (AEA).
Broadly speaking, the idea is that these AEAs will effectively become experts in all things hospitality, by trading, organising and analysing data on behalf of the individuals or the organisations that own them, to find and create value from previously siloed data.
So, is Fetch just another travel blockchain start up? Not according to Sheikh. “Blockchain is a very generalised term. I prefer to call it decentralised ledger technology (DLT),” he says.
Also, ‘blockchain’ is just a part of what Fetch is aiming to deliver. Instead, it is honing in on two things: AI and machine learning on the one hand, and decentralisation using DLT/blockchain on the other.
The first part – AI and machine learning and the potential to deliver – is already fairly well understood by the travel industry. However, the decentralised piece involving blockchain/DLT, says Sheikh, “is a bit more difficult for them [the travel industry] to absorb”.
If you are asking a big corporate to change the way they are doing things today to become [purely] blockchain-based, then that isn’t going to happen
The reason, as many well know, is the presence of huge internal centralised management and integration systems. And on one thing Sheikh is clear: “If you are asking a big corporate to change the way they are doing things today to become [purely] blockchain-based, then that isn’t going to happen”.
However, Fetch isn’t asking the industry to make a huge leap. Instead, Sheikh explains, it’s about building a bridge; a transition stage that inexpensively connects and integrates existing, often siloed, centralised internal systems with next-generation decentralised ones. After all, companies may have those central internal systems but they also are already interacting with customers via largely decentralised web-based platforms. Simply put, blockchain just adds to that system.
“People can say what they like, but in terms of the transaction a blockchain-based system is more stable, more secure, more auditable. That is a proven fact,” stresses Sheikh.
As Joerg Esser, a former Thomas Cook group director, theoretical physicist and now consultant to Roland Berger, pointed out at EyeforTravel Europe earlier this year, “blockchain enables trust at scale”.
Independent and trustworthy
In principle, that all sounds great but, and this is a big but, for ‘decentralisation’ and the commoditisation and democratisation of AI to succeed, Sheikh acknowledges that “there has to be an independent trust mechanism”.Today this ‘trust mechanism’ between transacting parties is delivered by middlemen like Airbnb or Booking.com, and it comes with a price tag.
With its AEA, Fetch is looking to change that for the travel industry. “If you can have an independent, decentralised system which everybody knows that nobody can fiddle with and the ratings that are provided are true, then people will start to trust it,” he says.
Better still, this will be good for the bottom line, because the likes of Airbnb incur a cost to the consumer of +-10% versus a decentralised system with a cost of 0.1%. “That’s a huge difference!” Sheikh adds.
AEAs will provide that bridge, and “you don’t need to spend millions of dollars changing your old system into a new one. You have something sitting in the middle – an agent – that can do that for you.”
AEAs: the potential use cases
1. Relevant offers
At the simple end of the spectrum, said the press release following Fetch-wi-Q partnership announcement, “the agent might upsell a late checkout based on the amount of alcohol ordered by the guest a day earlier. However, it’s the ability of the autonomous agent to integrate contextual understanding from the wider Fetch system that makes advanced uses possible. For example, if there is a major sporting or cultural event occurring in the destination, the agent will offer products it knows people are more likely to select, for example a halal meal during Hajj, or perhaps burgers during football tournaments.”
2. Rethinking the role of the room night
Another application could be in the hospitality space where room night bookings today are handled by reservation systems, and users still go online to find the right room using platforms of search and discovery, most notably Google. However, even if the search giant – itself a decentralised system – is the facilitator, it’s still not autonomous. As Sheikh puts it: “Travellers still have to go online and input what they are looking for, and that creates a lot of friction. It can be a nightmare.”
This, he argues, is not the future. Instead, the idea is that if the AEA will be clever enough to use the right machine learning and AI, in order to find and book the optimum room, so that users don’t need to be engaged in the process.
By taking such a bottom up approach, Sheikh argues that it is possible to radically reduce the integration costs associated with massive room reservation systems. In essence, the room itself becomes its own self-contained agent, which looks after its own things. So, if a television is broken, the room’s AEA would report that rather than someone having to check it out first. “There will be always be a physical element but as the electronics become more and more sophisticated, the agent will receive reports and find even better solutions quicker,” he explains.
New business models
In a Fetch-styled world, current business models don’t apply. The analogy, Sheikh argues, is this: “Who owns the internet? Who is the supplier of the Internet? There isn’t one and so nobody makes money by just supplying this network. You then make money by supplying services on top of this.”
In tech terms, the internet of today has a thin protocol layer, with a fat layer of applications sitting on top; data today sits in the application layer and is controlled by the many not the few. The blockchain stack with its ‘fat protocol’ layer and thin applications’ layer reverses this. Still confused? This Coin Crunch video offers an explanation.
What is becoming increasingly clear is that more data means better AI, but the fact remains that the current status quo is not delivering. This lies at the core of the ‘democratisation structure’ that Fetch is proposing. It wants to put control of data back into the hands of the people by providing them with an autonomous economic agent. “This is your agent, your data and nobody has access to it,” says Sheikh.
The better the prediction you make, the more value you can unlock
Simply put, Fetch wants to build models that interact with each other and learn from those interactions. In essence, this is about predictions, which people make all the time based on context, on preferences, and on the data to hand. Importantly, as Sheikh points out, “the better the prediction you make, the more value you can unlock. And if we can improve search and discovery and enable integration with the outside world in a cost efficient manner, then I don’t see why people wouldn’t take up such a service.”
Clearly, the Cambridge-based company is nothing if not ambitious! But just how far off is a world where we control our own data, and are supported by our very own friendly, fully autonomous agents to end the nightmare that is travel booking?
Fetch is currently working with various corporates to develop an AEA, including one for the healthcare industry, though no names can be mentioned yet. The first, however, are only expected to be fully deployable by the end of next year.