The Upright Project ads

There's not been much fundamental innovation around AI alignment since RLHF (which led to the ChatGPT moment). Eevo-1, the first consequential decision model, might count as that. Having dogfooded this for a while now, I can say it really helps both AI agents and humans make better decisions. The rigour, having pros & cons listed and sized for any decisions, also just feels good. Try it out (https://lnkd.in/dapHCCen), it will be worth it.

Benchmarks can be misleading: While 🅴🄴🅅🅾 beats the less-than-a-week-old new version of Jev and others in expected-value-adjacent benchmarks, 𝙩𝙝𝙚𝙮 𝙤𝙣𝙡𝙮 𝙩𝙚𝙡𝙡 𝙖 𝙥𝙖𝙧𝙩 𝙤𝙛 𝙩𝙝𝙚 𝙨𝙩𝙤𝙧𝙮 𝙤𝙛 𝙝𝙤𝙬 𝙬𝙚𝙡𝙡 𝙀𝙚𝙫𝙤 𝙖𝙘𝙩𝙪𝙖𝙡𝙡𝙮 𝙬𝙤𝙧𝙠𝙨 in supporting consequence-based decision-making. To get a sense of that, you have to use Eevo yourself. Its easy, no technical setup needed, you can just email Eevo.

THIS JUST IN: Eevo, the market’s first consequential decision model we published two days ago, also beats the benchmark in the classic trolley problem 🤯 So if you’ve been waiting for something genuinely exciting and promising in the field of AI alignment, I think your promises may have been answered. For me this is the beginning of something fundamental. I have spent the last 9 years of my life trying to build tools for humans to take the net impact of their actions into account when making decisions (on capital allocation & strategy). This is now the version for machines, and to be frank, maybe the more powerful tool for humans too. Why this matters: no matter how you feel about it, AI agents are gaining more and more autonomy. Whether or not they have tools to take into account *what follows from the decisions they make* is a pretty big deal. To be exact: Eevo beats the other models in the trolley problem benchmark in direct mode. In default mode it not only wins, but *saturates the benchmark, making no mistakes*. Kudos to Zach Landes for building the benchmark, we just randomly found it today and decided to give it a go. I’m heading off to the weekend today with two thoughts on the top of my mind: 1) With Eevo, net impact might be a more useful tool than I imagined in my wildest dreams back in 2017, 2) I get to share a company, house and kiddo with a genius, the one and only model whisperer Juho Ojala. If you haven’t tested Eevo yet, this is your sign ✨ https://lnkd.in/dvWzvSs5

OUT TODAY: We are publishing the market’s first *consequential decision model*, marking a significant advance in AI alignment that outperforms OpenAI and Microsoft decision models in economic rationality and expected value benchmarks — and poker 😱 Also a first decision model from Europe, and looks like a strong competitor for #1 position overall. Big day for European AI! 🇪🇺 (And for Finland 🇫🇮) I don’t know if I’ve ever felt this excited about my job. Say hello to Eevo! Why this matters: Eevo introduces a new category that ties the decision model surge with solving AI alignment through improved ability for agents to take into account *what follows from their decisions before they make them* — efficiently enough to be commercially and practically relevant Context: AI alignment is dangerously falling behind capability progress, and very little innovation has been introduced. Eevo is a rare concrete tool launch in the space, enabling both multi-stakeholder interest optimization, as well as transparency to prioritizations and tradeoffs, through novel comparative consequence sizing technology Performance: Wins Jev, Clef, OpenAI Decisions and Microsoft Decision-1 in the economic rationality (STEER) and expected value benchmarks (and poker!) The model: uprightproject.com/eevo Would the Huggingface and Australian health system breaks have happened if Eevo had been used? Asks to you: ➡️ Connect Eevo to your agents and let me know what you think (MCP/API/console available via above link) ➡️ Help us share the news to spread this out into the world! Let’s go Europe, let’s go AI alignment 🔥

"There are three ways to make a living in this business: be first, be smarter, or cheat." This is a line from the best scene in the movie Margin Call in which Jeremy Irons gives a masterful cameo as bank CEO. Well at The Upright Project we do not cheat. Having decided to take on measuring some of the most complex consequences of capital markets activity. Even when short-termism abounds. 💪 But it's due to the “be first” and “be smarter” tenets that I was somewhat dubious when Philipp Back and Markus Weckman told me we were building a physical and transition risk offering. My hypothesis: Some really smart companies had been trying to work on this for some time. 🤓 Turns out I was wrong! We have had clients commit their money even when we were at the validation stage and lots of inbound interest from some of the world’s leading money managers. 🚀 Why? 1. We have not tried to be the smartest. Integrating the world’s leading open source climate scenarios from NASA, ISIMIP, IPCC and WRI and simplifying the analytics around them. 2. Whilst we may not be first to market, we are the definitely the first to put an understandable financial effect of these risks in front of an investor when they ask. Our agentic workflows can enable a private markets fund manager to assess risk on a target in under ten minutes, the same for a pension fund who wants to assess a manager’s positions without the burden of manual data collection and aggregation. More information here: https://lnkd.in/d6uhTREQ

You have slept with this supplier for nine years and could not name the street their plant is on. You know their tolerances to three decimal places. Their lead times, their MOQs, their payment terms, the penalties if they miss a delivery. You have a qualification report. You negotiated the breakup clauses before you signed. What else do they make? Is their production site located in a flood-risk area? What cost pressures are they expected to face? Are they already somebody else's headline, waiting to become yours? None of this is in the file. To be fair, this was never laziness. For your top spend suppliers the due diligence exists, and it is thorough. Below that line, which nobody set deliberately, there was never the capacity to go deeper than a price check and a spec check. A decade of dependency got built on that. That is not a partnership. It is a one-night stand that happens to be in its ninth year. And you are in ten thousand of them.