The brief history of artificial intelligence: The world has changed fast – what might be next? Just 10 years ago, no machine could reliably provide language or image recognition at a human level. But, as the chart shows, AI systems have become steadily more capable and are now beating humans in tests in all these domains.

The brief history of artificial intelligence: The world has changed fast – what might be next? Just 10 years ago, no machine could reliably provide language or image recognition at a human level. But, as the chart shows, AI systems have become steadily more capable and are now beating humans in tests in all these domains.

Training computation is measured in floating point operations, or FLOP for short. One FLOP is equivalent to one addition, subtraction, multiplication, or division of two decimal numbers. All AI systems that rely on machine learning need to be trained, and in these systems training computation is one of the three fundamental factors that are driving the capabilities of the system. The other two factors are the algorithms and the input data used for the training.

For the first six decades, training computation increased in line with Moore’s Law, doubling roughly every 20 months. Since about 2010 this exponential growth has sped up further, to a doubling time of just about 6 months. In her latest update, Cotra estimated a 50% probability that such “transformative AI” will be developed by the year 2040, less than two decades from now. Many AI experts believe that there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner.

On the other hand, let's talk about "The red flags and magic numbers that investors look for in your startup's metrics - 80 slide deck included!" At the moment that your New+Reactivated is equal to your Inactive users, each time period, then you hit peak MAUs. This is the thing to watch for, because then it’s all flat or down from there. The problem is that the Growth Accounting Framework provides for lagging metrics. It’s hard to predict the future.

To think about the quality of the loops – how defensible and proprietary are they? How scalable and repeatable? Is there upside in optimizing them or adding to them further? The key thing to ask for the Acquisition Loop is to understand how a cohort of new users leads to another set of new users. If I see a startup that doesn’t directly ask for the signup, I assume there’s upside that can be gained. Minimum design is better.

The first thing to ask for is the product’s Acquisition Mix. This is a look at signups broken down by channels/loops and by time period (ideally weeks). I’m looking for signals that the dominant channel(s) are proprietary and repeatable. Ideally they are loops. It turns out that one of the biggest determinants of “quality” of new users is the source of the user. As a result, you want to understand both how signups are being generated by various channels, via the Acquisition Mix report above, but also a sense of the quality by understanding the activation rate by channel. If all the users have come from beta users list or Product Hunt, that won’t scale over time. On the other hand, if marketing spend and product efforts are going towards high-quality channels, that’s fantastic.

It’s important to understand the underlying platform of any acquisition loop because things can collapse quickly. Once you have all of this together, then you ought to be able to create a series of scenarios on where your growth curves are going to go. If you have a network-based product, like Dropbox or Slack, then you need active users to engage each other. If it’s purely a utility, then you want engagement in one time period to help set up engagement in a future time period.

There are linear channels to re-engage users. These are useful, of course, but again, they don’t scale. It’s better when users re-engage each other or when users re-engage themselves. The social feedback loop fundamentally is built on the content creation step. If it’s not easy, then it won’t work. So it has to be an activity that a lot of users want to do. One key aspect of every network is the density of connections. It’s important to build the number of connections up, but they have to be relevant. The cohort curves need to flatten. Ideally >20%, so that each signup activates into a sticky, active user over time. To detect artificial engagement that’s being manufactured, not organically created by users, you can look at a breakdown of every notification that a product sends out. And the volume and CTRs over time.

What I want to understand with a Frequency diagram is to segment high- and low- frequency segments, and start digging into their usage of the product. If you can upsell new use cases, then there’s a ton of upside. Engagement metrics are very hard to move compared to Acquisition. As a result, it’s better to assume the curves are what they are. But if you must add a bullish forecast, the right way to go is to focus on new user activation. And up-selling users from one frequency segment into the other. This is why achieving network density and easy content creation is so important- you need ways to bring people back into the network.

In conclusion, as AI systems continue to advance and exceed human capabilities, the development of human-level artificial intelligence within the next few decades seems increasingly likely. The exponential growth of training computation, along with algorithms and input data, plays a crucial role in driving the capabilities of AI systems.

On the other hand, for startups, understanding the key metrics and loops that drive growth is essential for success. By analyzing the acquisition loops, the quality of new users, and the engagement metrics, startups can optimize their strategies and focus on areas that have the potential for scalability and repeatable growth.

To achieve growth and sustainability, it is important to create proprietary and defensible loops, ensure easy content creation, and build a dense network of relevant connections. Additionally, focusing on new user activation and up-selling users from one frequency segment to another can drive engagement and expand the user base.

In summary, for the future of AI and startups, continuous innovation and optimization are key. By harnessing the power of AI and understanding the metrics that drive growth, we can navigate the ever-changing landscape and unlock new possibilities.

Actionable Advice:

1. Focus on developing proprietary and repeatable acquisition loops to drive growth.
2. Prioritize easy content creation to encourage organic engagement and user-generated content.
3. Build a dense network of relevant connections to enhance the user experience and drive long-term engagement.