Nov 11, 2022
Machine Learning versus Human Decisions: Which is better
As a platform that aims to provide fast accurate valuations for rare non-fungible tokens (NFTs), Lithium Finance’s algorithm uses a hybrid pricing model that combines both machine learning and human inputs for the valuation process.
Since machines crunch data much faster than humans, why not let computers take care of it? It is true that machines process quantitative models with a large number of variables efficiently, but they stumble in certain situations that would seem logical to humans. For example, humans understand that ice cream sales and the number of people getting sunburnt while at the beach increases on a sunny day. On the other hand, artificial intelligence may draw the conclusion that eating more ice cream increases the risk of getting sunburn. Machine learning correlates the two events even though ice cream does not cause sunburn.
To understand this idea further, we need to explain both machine learning and human decisions further.
What is Machine Learning?
Machine learning is a method of data analysis that systems use to ‘learn’ from data, and the goal of a machine learning system is to predict or classify an output based on a given input. By utilizing machine learning to make sense of large data quantities, we can create a predictive output in a short amount of time. Some examples of machine learning we are familiar with include email spam detection in our email inboxes, and recommendations when you browse e-commerce or streaming platforms.
What is Collective Intelligence?
Collective intelligence is the combined intelligence stemming from the conscious collaborative efforts of many individual humans making up a group, and appears in consensus decision making. We are most familiar with this particular intelligence when we work as a group to complete a project, and a digitalized example is sourcing product reviews through a rating system.
What do they both lack?
Machine Learning
Going back to the example at the start of the article, we established that machine learning isn’t the solution to everything. While it has the ability to analyze complex data sets and search for patterns which the human mind may not be able to visualize, it needs a relatively large set of data to produce a reliable input.
However, each NFT is unique with different traits, especially the rarer ones that are never traded after minting. Machines can make predictions from data on comparable items, but it cannot produce price prediction without sufficient historical data. Rare NFTs don’t change hands frequently, limiting the historical data for Machine Learning to work on. Therefore we can see that machine learning alone cannot account for an accurate valuation in some situations.
Collective Intelligence
Trading is a complex economic human behavior, and who understands and predicts these behaviors better? While machines can be trained to understand human behaviors, it takes a lot of data to avoid situations like misinterpreting sunburns as being caused by ice creams. Imagine the process and data required to train a computer vision system to recognise facial expressions; the same is a lot more efficient and almost effortless for a toddler in contrast.
It takes a human to recognize human behaviors especially when we are dealing with limited data, and they can put irrational market sentiment into the correct context when they come across outliers or changes in market conditions. The downside is, the process is effortful and often time-consuming.
Why Machine + Human?
By understanding the pros and cons of machine learning and collective intelligence, we can see that a combined approach where human appraisers interpret and fine-tune the outcome of quantitative models, while machine learning aids humans in getting a better understanding of complex relationships, is more reliable.
This is particularly so when we are dealing with price predictions for illiquid assets like rare NFTs with limited trading data, which are decisions made through human behaviors. When insufficient data is available to machine learning to produce reliable predictions, we can help algorithms make better sense of human decisions by supplementing machines with expert insights from the community.
Now that you know our foundation of having Collective Intelligence and Machine Learning work together, we will look at how Lithium works in the next chapter! Stay Tuned!
About Lithium
Lithium Finance is the first decentralized NFT valuation protocol powered by collective intelligence and machine learning. Redefining NFT valuation approach through incentivizing honest assessment from community to reveal market sentiments.
