Weekend Reading #215

This is the two-hundred-and-fifteenth weekly edition of our newsletter, Weekend Reading, sent out on Saturday 29th April 2023.

To receive a copy each week directly into your inbox, sign up here.

*****

What we’re thinking

All hail the greatest entrepreneur of his generation. No not Bezos, not Elon, but Mark Zuckerberg. Is he successfully reinventing his company yet again? After his much-maligned foray into the Metaverse, his stock crashed. We have written many times of our admiration for him as an entrepreneur and how it's best not to write him off and here is why. The thing about Zuck is that he is young. He is not yet even 40 years old. He is young enough to learn from his mistakes and fix them. 18 months after betting the farm on the Metaverse and even changing the company’s name. 18 months of earnings disappointments and the Metaverse seemingly disappearing into the ether (yes pun intended). This guy rolled up his sleeves, chopped 10 000 jobs, and went again.

He has also learned to be better with words as he introduced today’s hot topic right alongside the ongoing commitment to the buzzword of “yesteryear”.

“A narrative has developed that we’re somehow moving away from focusing on the metaverse vision, so I just want to say up front that that’s not accurate,”

“We’ve been focusing on AI and the metaverse, and we will continue to.”

Got to hand it to him. New year, new story, new set of numbers and his stock is up 150% so far YTD. He has balls of steel and once again we caution on betting against him. While Elon is saving the world, he is saving his business.

Eugene and Harry will be in Singapore and Hong Kong for the next couple of months. If you are based in either of these locations and are keen to meet up, please let us know!

What we’re doing

The past few weeks have been jam-packed with meetings and preparation as we look to move forward with a round of fundraising for our internally developed software business. We started the week with our heads down as we powered through the final iterations of our financial model and investor presentation. We then spent the second half of the week with our newly appointed CTO, Stav, who flew into London specially to help us make a detailed plan for the coming 18 months. Stav has been deeply involved since the very beginning and has been instrumental in putting together the systems architecture that underpins our offering, and we are extremely excited to have him on board in an executive position. We talked through both our business and technical strategy, and even our hiring plan as we scale up the business and hire more developers. We’ve made some huge strides recently with contracts signed and connections set up.  Things are likely set to accelerate as we head to Asia to bring on new counterparties and grow the business to the next level. HS

Having recently been admitted as a member of the FIX Trading Community, we were welcomed with our official initiation call taking place this week. For those that don’t know, FIX (or Financial Information eXchange) is a technical protocol which governs communications and trade messages across financial services firms. Whilst membership to this group in itself may not sound like much, we find ourselves in good company when looking at our fellow community members. Our first official engagement as members will be at the FIX APAC Regional Conference early next month in Hong Kong, and in the coming months we’ll be working closely with some of the organisation’s working groups in determining the future of the protocol. HS

After hearing and reading so much about Auto-GPT, I thought it’d be a good idea to give it a shot and try to set it up. The idea of a master GPT instance that spawns subordinate GPT instances recursively to complete tasks is certainly fascinating and in theory could lead to an infinite level of reasoning, but there was no other way to find out than to actually deploy an instance of it. So, I spun up a brand new shiny Microsoft Virtual Machine instance on Azure (where better to try GPT deployments than on “home” territory) and after a couple of hours getting everything from Python to pip to conda and git set up, got Auto-GPT up and running.

The outcome? I’ll caveat this by clarifying that I’m by no means a devops expert (far from it), nor a machine learning aficionado, so maybe the set up isn’t as conducive as it needs to be for Auto-GPT to really shine, but the first few runs of Auto-GPT were a little bit underwhelming and disappointing. The concept, however, is impressive and opens the doors to a MUCH deeper rabbit hole – namely, what happens when GPT is no longer confined to a “chat” UI (as with ChatGPT) but is actually given control of your command line interface. Plugged into a live server, Window or Linux alike, at the command line level with administrative privileges, a GPT instance can actually make quite an effect (or do quite a lot of damage). Auto-GPT has a kill switch in that the proposed commands are first shown to the user before they are executed – but what if we took that kill switch off? Efficiency, yes, but also it becomes nearly unstoppable.

To be fair, I’m by no means saying that the rave reviews about Auto-GPT are false. Everything Auto-GPT is said to be able to do, it can truly do. Perhaps it was just a little bit too much to ask it to basically take on the role of a junior analyst for a hedge fund... Or is it just a matter of more iterations and training? After all, the prospect of being able to make demands that are not “humanly” possible, demanding a level of perfection that is “inhuman”, of a junior analyst simply because said analyst is now no longer human is extremely appealing. EL

What we’re reading

Ray Dalio put out a piece this week called “What I Think Is Going On 1) With China-US relations 2) With their relations with other countries 3) In China”. He published it on Linkedin which for some reason is his platform of choice. It is genuinely a superb summary of US-China relations, its complexities and most importantly the trigger points for disagreement. He makes some proposals to reduce inflammation, but despite an obligatory hopeful paragraph at the end it sounds to me he is resigned to the fact that bad times are ahead. He knows where this is heading sadly better than anyone. Some say war is already underway in space. Let's just hope it stays there. DC

What we’re watching

Readers of this publication will be used to us writing about Stan Druckenmiller, in our view the greatest public markets investor there has ever been.  He hadn’t popped up in an interview for a while but this week he appeared on video to speak at Norges Bank’s annual investment conference.  He repeated his view that the current environment has no historical parallels, and his conviction is really low.  His one high conviction trade is the short the US Dollar.  The interview is a short one but full of good stuff.  Most enjoyable as always to me was how he gave insights into his trading process. So much to take away. DC

What we’re listening to

Continuing my search for AI enlightenment I listened to the 2021 BBC Reith Lectures featuring Stuart Russell, Professor of Computer Science at the University of California, Berkeley amongst other titles.  Living with Artificial Intelligence is the name of the series, and the lectures are four in total – each covering a different subsegment of how he thinks AI will influence the world.  They cover warfare, the economy and what the future holds for humans.  After listening to a lot of wild takes in recent weeks and months, this one was somewhat more centred.  Unfortunately, the conclusions are same.

Speaking of centred, Bankless hosted the other side of Eliezer Yudkowsky this past week in Paul Christiano.  Yudkowsky himself recommended him to Bankless when asked who they could speak with to get the other side of the bleak outlook he presented to them all those weeks ago.  Christiano is one of the most respected voices on AI alignment and is known for running the language model alignment team at OpenAI.  His interview is rather less controversial, and he presents himself rather more softly than Yudkowsky but even though he is much more optimistic, he still has rather high odds of adverse outcomes.  As one of the “optimists” this was surprising to me.  His primary difference with Yudkowsky was in fact the speed at which AGI would develop.  Whereas Yudkowsky believes once we cross that threshold it will be almost instant, Christiano sees a far slower uptake giving us all time to react and adjust.  Somehow this didn’t make me feel any better.  A reader recently recommended listening to David Deutsch who by far of all the takes, presents the least frightening scenario and comes also with great pedigree.

On a lighter note, I’ve been listening to some good electronic music recently and a good friend recommended an Aussie artist called Flume, which sounds far better than his real name of Harley Edward Streten.  Pretty cool stuff.  Give him a listen if you are into that kind of music or even like me just keen to see what else is out there. DC