We are on the verge of a super bubble created by AI due to the injection of trillions of dollars to support AI innovation.
Here is a cheat sheet to what projects to look in AI as an investor
1/25 ๐งต๐๐ผ
Here is a cheat sheet to what projects to look in AI as an investor
1/25 ๐งต๐๐ผ
2/ Have your popcorns handy, we are in for a ride.
- Why consider AI for investment?
- What is AI and how does it work?
- Sector wise analysis
- Final thoughts
- Why consider AI for investment?
- What is AI and how does it work?
- Sector wise analysis
- Final thoughts
3/ Why should we even consider AI as a sector to invest in?
Increased funding for web3 AI projects as compared to previous years
Increased funding for web3 AI projects as compared to previous years
4/ Stunning growth in the real world and funding in 2023
5/ Funding Winter didn't affect AI companies as much as it did others
6/ Peak Attention
7/ What exactly is AI, and what's the behind-the-scenes process investors should get familiar with?
Imagine AI as a smart robot that can think and perform jobs just like people do.
But how do we bring such a marvel to life?
Imagine you're crafting an essay. You'd search online, gather info, and write. Building an AI to write essays follows a similar, still more technical path.
- Data-Collection: Gather an extensive library of top-notch essays and a wealth of related information.
- Model Creation: Next, build a computer program that learns from these essays to write new ones.
- Training: Teach our model the art of essay writing using the treasure trove of essays and information we've amassed
- Inference: Finally, we put our creation to the test: we ask it to produce an essay on a completely new topic
Imagine AI as a smart robot that can think and perform jobs just like people do.
But how do we bring such a marvel to life?
Imagine you're crafting an essay. You'd search online, gather info, and write. Building an AI to write essays follows a similar, still more technical path.
- Data-Collection: Gather an extensive library of top-notch essays and a wealth of related information.
- Model Creation: Next, build a computer program that learns from these essays to write new ones.
- Training: Teach our model the art of essay writing using the treasure trove of essays and information we've amassed
- Inference: Finally, we put our creation to the test: we ask it to produce an essay on a completely new topic
8/ As an investor in the field of AI, where should your focus lie?
1. AI Training Data Set/Storage
2. Computational power
3. Marketplace For AI Models/Engineers
4. Apps/Infra building on top of successful models
Let me explain
1. AI Training Data Set/Storage
2. Computational power
3. Marketplace For AI Models/Engineers
4. Apps/Infra building on top of successful models
Let me explain
9/ Money Making Machine 1: AI Training Data, Set/Storage
- It's noted from several studies that "AI models are only as good as the data theyโre trained on"
- High-quality data[correct and vastness of data] ensures accuracy while training on low-quality data would result in the AI machine being biased or outright wrong in many cases
- It's noted from several studies that "AI models are only as good as the data theyโre trained on"
- High-quality data[correct and vastness of data] ensures accuracy while training on low-quality data would result in the AI machine being biased or outright wrong in many cases
10/ The market cap of AI training data is projected to grow at a CAGR of 23.85%
As AI gets better, everyone wants better data, making it more valuable. Right now, the big tech companies are leading by buying smaller ones for their data.
As AI gets better, everyone wants better data, making it more valuable. Right now, the big tech companies are leading by buying smaller ones for their data.
11/ Fear of concentration of power in the hands of the bigtech could lead some towards decentralized protocols working on the same problem.
But still how efficient and useful data they can provide compared to google, amazon and microsoft who invest billions in collecting data worldwide is questionable.
Increase in smartness of AI models will also increase the demand for decentralized storage of these high quality data so that they can be accessed by anyone and it's not gated.
But still how efficient and useful data they can provide compared to google, amazon and microsoft who invest billions in collecting data worldwide is questionable.
Increase in smartness of AI models will also increase the demand for decentralized storage of these high quality data so that they can be accessed by anyone and it's not gated.
12/ Money Making Machine 2: Computational power
We need GPUs to train the AI machine with the data we have collected and this need for GPUs is set to increase exponentially due to innovative models.
We need GPUs to train the AI machine with the data we have collected and this need for GPUs is set to increase exponentially due to innovative models.
13/ Although big tech controls the market still can't meet the the rising demand.
Hereโs the shocker :
"Compute requirements for machine learning (ML) training which have grown 10x every 18 months since 2010. In contrast, computing power is believed to have grown 2x during the period and the primary reason for this is the โlong lead time to build new supplyโ
Heavy underutilization of GPUs from datacenters and idle ones sitting in homes also remains the reason why the supply has not met with the demand.
Hereโs the shocker :
"Compute requirements for machine learning (ML) training which have grown 10x every 18 months since 2010. In contrast, computing power is believed to have grown 2x during the period and the primary reason for this is the โlong lead time to build new supplyโ
Heavy underutilization of GPUs from datacenters and idle ones sitting in homes also remains the reason why the supply has not met with the demand.
14/ This presents a golden opportunity for web3 protocols in the Distributed computing space as they try to utilize the underutilized and consumer-held/sitting-idle GPU to effective use.
Due to the nature of this business model where underutilized compute is aggregated and used, these protocols are able to provide computational resources at a lower cost with good speed, which is a huge USP imo.
Due to the nature of this business model where underutilized compute is aggregated and used, these protocols are able to provide computational resources at a lower cost with good speed, which is a huge USP imo.
15/ Some protocols
16/ Money Making Machine 3: Marketplace For AI Models/Engineers
Building a top tier AI model from the ground up costs a fortune, easily hitting the million-dollar mark and requires extremely skilled developers who are hard to hire
Building a top tier AI model from the ground up costs a fortune, easily hitting the million-dollar mark and requires extremely skilled developers who are hard to hire
17/
Most of the companies successful in AI are closed source[such as GPT3,4, Claude, etc] and are well funded.
But this presents a great problem where the barrier of entry of a new guy is so high that it normalizes the monopolistic nature.
Although there are a lot of open source models being created to compete with the closed source, the feasibility of these models are still a question.
Most of the companies successful in AI are closed source[such as GPT3,4, Claude, etc] and are well funded.
But this presents a great problem where the barrier of entry of a new guy is so high that it normalizes the monopolistic nature.
Although there are a lot of open source models being created to compete with the closed source, the feasibility of these models are still a question.
![17/
Most of the companies successful in AI are closed source[such as GPT3,4, Claude, etc] and are...](https://pbs.twimg.com/media/GGn49XoWMAAw6oW.jpg)
18/ The above concerns put a lot of focus on decentralized AI model marketplace where anyone can use a trained model without having to spend a fortune to train it and trainer getting paid whenever someone uses it all in a decentralized manner.
Still the feasibility of it is unknown as the model keeps getting bigger and bigger.
Still the feasibility of it is unknown as the model keeps getting bigger and bigger.
19/ Money Making Machine 4: Apps/Infra building on top of successful models
99% of the worldโs population will not tinker with the AI model, they will only interact with the applications created on top of these models.
Apps built on AI, like chatGPT, quickly attracted 100 million users, showing the massive appeal of AI-powered applications.
Text and generative apps built on AI models have seen huge user interest but are still searching for lasting revenue strategies.
Meanwhile, the development tools for these apps, like LangChain, BabyAGI, and HuggingFace, have quickly become popular, showing the fast-growing demand for AI innovations
99% of the worldโs population will not tinker with the AI model, they will only interact with the applications created on top of these models.
Apps built on AI, like chatGPT, quickly attracted 100 million users, showing the massive appeal of AI-powered applications.
Text and generative apps built on AI models have seen huge user interest but are still searching for lasting revenue strategies.
Meanwhile, the development tools for these apps, like LangChain, BabyAGI, and HuggingFace, have quickly become popular, showing the fast-growing demand for AI innovations
20/ Using AI to optimize Defi yields, to conduct through security audits, bots have been the major use case but this will change when the lines are blurred between web2 and web3.
Impact of user generated content created using AI [AIGC] will be high such as user generated NPC in games
Also monitoring the protocols that are working on bringing transparency to AI by using zero knowledge proof to record all the execution on chain would be worth the time.
Impact of user generated content created using AI [AIGC] will be high such as user generated NPC in games
Also monitoring the protocols that are working on bringing transparency to AI by using zero knowledge proof to record all the execution on chain would be worth the time.
Final thoughts :
1. The state of AI is in its absolute infancy. So splitting the industry into categories and backing the leaders or unique players could lead to a jackpot
2. Right now, AI is like a treasure hunt where companies spend a lot to strike gold, and the rules of the game are still being written. So, investing wisely is key.
3. Keep an eye on: Computing Costs: These could soar if global events like wars or sanctions hit supply chains.
Regulations: While major roadblocks to AI development are unlikely, staying alert to new laws is smart.
4. Being able to understand technical papers is crucial to avoid sinking money into vapourware projects.
5. Web3 vs. Web2: The battle is on. Web3's edge wonโt just be about being decentralized anymore; it's about offering better services or prices than Web2 giants. Invest in those who can truly compete, and trade the rest strategically.
1. The state of AI is in its absolute infancy. So splitting the industry into categories and backing the leaders or unique players could lead to a jackpot
2. Right now, AI is like a treasure hunt where companies spend a lot to strike gold, and the rules of the game are still being written. So, investing wisely is key.
3. Keep an eye on: Computing Costs: These could soar if global events like wars or sanctions hit supply chains.
Regulations: While major roadblocks to AI development are unlikely, staying alert to new laws is smart.
4. Being able to understand technical papers is crucial to avoid sinking money into vapourware projects.
5. Web3 vs. Web2: The battle is on. Web3's edge wonโt just be about being decentralized anymore; it's about offering better services or prices than Web2 giants. Invest in those who can truly compete, and trade the rest strategically.
Loading suggestions...