What are you working on? Checking in with Josh Croft

Joshua Croft
Fetch.ai

--

I’m Josh; lead Autonomous Economic Agent developer and Head of Community, leading Fetch’s internal public relations team and our burgeoning team of AEA developers. Here’s a brief overview of what I’ve been working on and what I’m up to at the moment.

We’ve got a lot of things going on at Fetch and every day is different. I’d like to start by discussing some of the Autonomous Economic Agents we’re developing. There are many types of agents, of varying stages of complexity and development, but I’ll restrict myself to introducing two.

The first agent can be explained using a relatively trivial example but, for me, it shows exactly what Fetch will excel in enabling. I drink milk, not loads of it, enough to usually warrant a delivery a couple of times a week, but sometimes it’s just one delivery. The problem here is I forget to order, I like tea and I like my milk delivered. To get around this I have a scale under the milk bottle and when it weighs below 360g the agent on the scale (scale/Arduino/cell) sends a message to the milkman requesting he delivers some milk. With Fetch, the agent can find a cheaper milkman, can pay for the milk before delivery and get confirmation that the milkman will deliver. I like this: I get milk, I don’t have to worry about milk. The milk may be cheaper too.

I’m very excited by the second agent. Once finished it will form the most complete system like it in the UK and possibly the world.

Fetch have built a data structure in which every public travel service overlays every road in the UK. We’ve got gigabytes of ‘local’ business information, congestion and travel data so we’re able to simulate the impact of potential obstacles such as traffic in order to calculate optimal travel routes. It’s a phenomenal data set. Our agents allow users who are planning a journey to gain real-time information with meticulously detailed travel suggestions, provided by the concierge service of an agent booking everything for you.

Let me give you an example:

I’d like to go from Cambridge Airport to Glasgow at 7:45am. I’d like to use public transport for most of the journey and I’d like to go via Hull to visit the aquarium. I would like to rent a car in Hull, but I want to avoid traffic. I want to drive from Hull to Glasgow and stay in a hotel there, but if it makes more sense to stay overnight in Hull, so be it.

With our agents on the Fetch network, we can facilitate this request and an agent can understand it — in a prescribed manner. The Concierge Agent would talk to over 100 agents, and that number could be far larger for complex journeys. Your agent will use the Fetch network to create this journey with many other agents selling information or services to your agent.

Your Concierge Agent would first build a fastest route before engaging with any other agents. When the fastest route is known, the Concierge Agent would book/advise on taking a bus from Cambridge Airport to Cambridge North Train Station. The Train Network Agent could be called to book your train service, for example the train from Cambridge North Train Station to Peterborough and then Peterborough to Hull.

When at Hull, the Concierge Agent would already have booked you a hire car, from information you provided or from data generated when you hired cars before.

You arrived in Hull at 3:00pm and you’re now able to drive to the Aquarium. Your tickets have already been purchased by your agent and are on your phone waiting to be used.

… you get the idea!

Developing agents is taking up a fair amount of my time, but earlier this year I was also focussed on other tasks. Recently I finished the first revision of the iOS wallet — Okay, this is a mock up:

The tip of the iceberg

What we can’t show yet, but we will soon, is the novel approach we can use to interface with agents on the wallet and the customisation of your own agent on the device. This is really exciting for Fetch, as the launch agents will shine. I’ve also been busy developing the new Fetch.ai website, which we brought back in house to try and better show our vision. We’ve got big plans for it, and in the future we see it as a central hub to the Fetch team and our development resources. Finally, if you follow us on Telegram, you’ll be aware of Fetchbot, the reinforcement learning agent I have been been developing, to help answer simple questions and be part of our community — it’s got a great development plan ahead of it, and will be very useful for those wanting to interact with the test-net at a later date!

You’ll be able to meet me at Fetch meetups and conferences around the world. Yes I wear a few hats, but I just really enjoy having an impact across the Fetch system, love being able to code in lots of different languages and working with some really exceptional developers.

Of course, I also really want to make Fetch happen.

Josh

--

--

Joshua Croft
Fetch.ai

Lead Application developer at Fetch.AI. Also not bad at riding a bike.