#5 Deeptech Analysed - ChatGPT, the ultimate AI language model for all your needs & Lilium Jet the future of transportation with revolutionary eVTOL technology
What's happening this week? 12/12/2022 - 18/12/2022
ChatGPT, the ultimate AI language model for all your needs
By Lirone Samoun , Deeptech Expert
ChatGPT, the algorithm that has reached over a million users and drove the tech world crazy.
What is going on?
You may have heard about it. If you're in the tech world, it was hard to miss the big buzz generated by the new artificial intelligence algorithm that answers almost all your questions. A conversational chatbot created by OpenAI (a San Francisco-based company that counts Elon Musk among its founders). OpenAI creates machine learning models and slots them into different applications. Their most famous Machine Learning (or ML) model so far is called GPT3, a model that can understand and write natural language. A version of this model is the foundation for ChatGPT, an AI chatbot trained on a blend of text and code.
What does it mean?
Ask anything you can think of and the AI will be able to answer you, sometimes in great detail. Solve a mathematical equation, write a cover letter, write an email, ask questions about life or life issues or generate stories. These are just a few examples among hundreds of others. Moreover, ChatGPT has unique memory; it can remember earlier comments in a conversation and recount them to you. It chats like a human who could be in front of you.
First, the tech behind ChatGPT isn’t new. It’s based on “GPT-3.5” an upgraded version of GPT-3 which became available to the public many months ago. As a reminder, GPT is an AI that can generate any text similar to what a human could produce. The first step to create ChatGPT was to adjust GPT-3.5 for conversations. They literally had human AI trainers provide conversations in which they played both sides—the user and an AI assistant. With a model capable of generating answers similar to humans, they needed a way to tell the AI what a good/bad answer was. To solve that, they used again humans to rank randomly selected answers that ChatGPT was spitting out from best to worst. The rank was then used to train a second model they called the "reward" model. So far, there are two models:
a model that can answer questions like a human.
a model that can say how good/bad the answer was.
Finally, the last step was to train Reinforcement Learning model which is like dog training where a reward is given for a "good" behavior. Is the reward a piece of chicken? Not really. They used the score as a reward to train the model.
Have a model that generate a humanlike answer.
Have a model to score that answer.
Have model to learn from the score and re-adjust the answer until it is satisfactory
Repeat until it reaches the desired accuracy.
Why does it matter?
💸For markets: new business opportunities
Here’s a list of what GPT technology can do:
Nonfiction: dialogue, impersonation, news articles, summaries, teaching…
Professional: ads, copywriting, CV generation, emails, content marketing, team management
Code: generate and understand any code in different languages
Creativity: fiction, songs, humor, cooking recipes, kids stories…
Rational skills: Logic, mathematics, explanation of concepts, meaning of life…
Just as image generators based on prompts (Dalle, stable diffusion...) disrupt the world of graphic designers, advertising and photography, ChatGPT and its associated technologies disrupt a whole bunch of sectors and domains. Pay attention to new opportunities coming up whether its investment or job creation.
🧑🏿🤝🧑🏻For society: beware of the slow fall
What kind of society are we heading to? This is the question we have to ask ourselves. Today, we are more and more connected, we depend more and more on the internet and services like Google. Is our brain being led to atrophy little by little by lack of direct use of it? What about critical thinking? Algorithms like chatGPT have a lot of limits but we don't necessarily know it, at least, for a lambda person who doesn't necessarily understand all the issues. Are we going to rely more and more on the machine? To what detriment? What about intellectual property? Who owns it? Finally, job losses. This kind of system threat computer-based, non-routine cognitive jobs even more than robotics threats the workers. Some research estimates that around 40–50% of all jobs could be replaced in 15–20 years.
Here the time it took for certain system to reach 1 million users:
Netflix - 3.5 years
Airbnb - 2.5 years
Facebook - 10 months
Spotify - 5 months
Instagram- 2.5 months
iPhone - 74 days
ChatGPT - 5 days
ChatGPT is one of those rare moments in technology that will reshape everything going forward. We live in extraordinary times where you see the launch of a new type of model that completely changes the AI space. Recent models (DALLE-2, stable diffusion…) have demonstrated and shown promising results in terms of quality and ability to understand the prompt. The last few months have been fascinating in terms of content generation. It's like it’s exponential. Now ChatGPT allows you to generate anything you want from text. Where is the limit? GPT 4 is already announced. We can't wait to see the new advances. Are we one click away from generating a whole bunch of media without thinking too much?
Lilium Jet the future of transportation with revolutionary eVTOL technology
By Gael Amouyal, Deeptech Expert
What is going on?
Before focusing on Lilium Jet, let’s first look at what is an eVTOL. Like the acronyms is suggesting an eVTOL is an electric vertical takeoff and landing airplane. The applications for this new kind of revolutionary aircraft are endless like air taxis, private transport, cargo transport, air ambulance, deliveries, inspection, mapping, surveillance and special mission, among others. An eVTOL can be autonomous or piloted and their designs are futuristic, mixing propellers wings and turbofans. The goal is to develop and democratize the air space as cities and roads are more and more impracticable. For more info on AirTaxis in general please read my previous article (AirTaxis and UAVs will they fly among us?)
What does it mean?
Lilium Jet is targeting the air mobility sector to provide air taxis and private transport to shorten travel time between busy cities. The story began in 2015 when the half size prototype flew for the first time. Since then the company fully demonstrated the eVTOL capacities, successfully raise founds, went public and works hard on certification. Its design is unique and the engineering is out of the box, with its wing in canard configuration and its 36 electric tiltable turbofans positioned on the rear flap which allow fully thrust vector control. A maximum range of 280 km is expected at 280 km/h cruise speed with a capacity of 7 seaters.
Why does it matter?
💸For Markets: When is it the best time to go public?
As Lilium Jet is not yet certified and the production aircraft not yet ready, Lilium does not make any profit yet. However Lilium already have a long list of customers around the globe with a total of 603 aircrafts already ordered. The price for one unit is estimated to be around $4 millions. Lilium successfully raised $945 Million with a last round of $119 Million last November. Moreover the company went public in September 2021 with the ticker symbol LILM after joining an SPAC (special purpose acquisition company). It is a market mechanism where a shell corporation is listed on a stock exchange with the purpose of acquiring a private company, thus making it public without going through the traditional initial public offering process. However the LILM stock value has only decreased since its creation to reach $1.20 (-90%). In my point of view it is the perfect example of why DeepTech companies should not go public until they have a product to sell. Investors are looking mainly on short-term results, but as a DeepTech company it will takes time before the Lilium Jet will be ready for production and we may expect delays and unexpected challenges.
🧑🏿🤝🧑🏻For Society: Let’s travel together!
According to Lilium, a typical travel from Manhattan to downtown Philadelphia will take only one hour and ten minutes and will cost about $200. Whereas the same travel in high speed train takes about two hours and costs $150. Moreover if we compare with a regular taxi, it costs $280 for a two and half hour drive. This new way of transportation can really become a real alternative for short travel and allow better mobility between cities, at a relatively accessible price.
🔮What’s next? Certification and production
We are all waiting for the first ride on a Lilium Jet, however the certification toward FAA and EASA need to be completed. According to its CEO Daniel Wiegand, in 2025 the eVTOL will be fully certified and ready for production. Lilium is taking the certification process seriously however, I personally thing we will
see the Lilium Jet in the sky in 2026 timeframe, as the production line need to be build and certification takes always more time than originally scheduled.