openAI
The world is changing at an exponential pace. It’s getting faster and faster. Despite going from one crisis to another, humanity is going forward and technology remains key in our modern world. The latest trend is about chatGPT but artificial intelligence, robotics and automation are really at the heart of current innovation. I’ve already talked about quick ways to automate your work. Today we’re at the next level. The founders of this company have collectively raised $1 billion USD in 2015 then Microsoft itself has invested another one billion. Their mission : ensure that artificial general intelligence benefits all of humanity… As you saw in the title, we’ll talk about OpenAI.
Their motto sounds great, better than SkyNet I would say. Yes, when we talk about AI, we often think about Sci-Fi movies like Terminator or i-Robot. I think that’s why OpenAI is trying to provide AI as a benefit for everybody. AI should not be feared, AI is something great, something that will help us, it’s the next phase for humanity… Or is it something that will take all our jobs, control all our lives and reduce our creativity.
I don’t know so many AI companies but I guess you can find a ton of them. These days, we all heard about OpenAI. A company founded in the silicon valley, one member is the very famous Elon Musk (him again), even if I don’t think he’s really active in the decisions, he still earns money from it. Sam Altman is the CEO and he represents the company. They have this “opensource” mindset and their AI products can be tested for free with some limitations. They are on a mission : OpenAI Charter.
We can trust their commitment but their investors put a lot of money in the company and at some point, they will have to make money, a lot of money. Even bots have to eat.
The strategy seems simple and efficient. They have great AI tools, they publish API for third party apps to use and you’re charged for using it in your own app. For example, you have a chatbot on your website and the backend of your chatbot is using the API of openAI to discuss with your customers and find solutions to their problems. This sales strategy can work if the tools behind the API are really powerful. This year, the organization expects $200 million in revenue. Next year the forecast is $1 billion. The company may be worth $29 billion… How is it possible ? This is due to the hype around AI and the latest buzz word on the Internet : ChatGPT.
chatGPT
It’s not often that you have the feeling of being in the future when testing something. ChatGPT is one of these tools/products. Don’t even finish reading this blog post and go test it before the free trial is over => https://chat.openai.com/
hi chatGPT, please describe yourself because I'm too lazy to do it
I am ChatGPT, a large language model trained by OpenAI. I can understand and respond to natural language inputs, and can be used for a wide variety of tasks such as answering questions, generating text, and more. My knowledge cutoff is 2021, so I may not have the most up-to-date information. Is there anything else you would like to know?
Ok, you’re back ? Impressive isn’t it ? So now, you have a super buddy who knows everything until 2021. You can ask him to write your book, your blog posts or any written content. You can discuss with your super buddy to learn new things to go deeper in the details of any subject. ChatGPT can write source code of complex app, you can copy/paste your code and ask him what’s wrong with your code. What is really impressive is the conversation, you can really discuss with this thing and it will keep the context of what you’ve discussed earlier together. With ChatGPT we’re already in the future, the future of AI.
The possibility of the ChatGPT API seems to be endless. Mainly we think about ChatBot and virtual assistants. The next step will be to add voice to it so you’ll be able to ask him (or it, not sure) using your voice only. Probably some companies will plug this to their smartphone assistant because Today Siri and Google seem so limited compared to ChatGPT. I’m sure we’ll be able to customize his name and you’ll have your own Jarvis like Tony Stark.
Let’s go on the steps after this. We could probably grant access to ChatGPT to our digital life. It would be able to make phone calls for us as Google explained with Lamda in their last event. In the near future I would be able to plug the ChatGPT API to my smartphone and to this blog and I would just have to say : hey ChatGPT publish a blog post about fishing, please add a generated picture on the same topic. Voilà ! The content creators would become useless, just ask a bot to do it. Articles from famous journals could be generated, music and lyrics could be generated, even Art could be generated by AI. On top of fake news, we’ll have fake content or non-human generated content : synthetic content will flood on the Internet with these tools; Just my 2 cents.
We can already find a lot of tools around automation, some tools already have chatGPT plugins to add in your automated workflow. You can use variables from a customer form and chatGPT will reply by a nice email, you can enter in your Notions list an idea for a new Youtube video and chatGPT can generate the title, the full text and even the thumbnail.
how do you work ?
How does it work ? It’s based on machine learning, neural networks and… Let’s just ask to chatGPT, I’m in the future, I’m too lazy to search and write it myself :
how do you work ?
I work using a state-of-the-art deep learning technique called Transformer-based architecture. My training data consists of a large corpus of text, which allows me to generate responses that are similar in style and content to the text I was trained on. When you input a question or prompt, I use my internal model to generate a probability distribution over the possible outputs, and then I select the most likely response based on that distribution. I use this process to generate coherent and contextually relevant responses to a wide variety of inputs.
What is important here : “I was trained on”. First, chatGPT says “I” like it was conscious of itself. More importantly, it was trained on. So they gathered documents, books, internet websites or any content prior to 2021 and everything was labeled. With these labels on documents, chatGPT is able to understand what it is and it can use a big algorithm to provide answers to our question based on matching and probabilities. OpenAI said chatGPT is not perfect, there are some errors. You can find many threads on the Internet about funny wrong replies from chatGPT.
I’ve heard the training of chatGPT was done mainly in Kenya with people far below the minimum salary. For sure, ChatGPT is not perfect. The question about the content rights will arise. Who will own the content rights on what chatGPT will produce ? Is there an issue with the content rights used by chatGPT to produce the new content ? Maybe we should create AI to protect AI rights.
dall.E
Another impressive tool from OpenAI is Dall.E, what is it ? Just ask to chatGPT, this thing is here to help me in the modern world.
what is dall.E ?
DALL-E is a 12-billion parameter version of the GPT-3 model, also developed by OpenAI. Like GPT-3, DALL-E is a state-of-the-art language model that can generate high-quality text, but DALL-E is unique in that it can also generate images from textual descriptions. For example, given a textual prompt such as “a three-legged dog playing a banjo,” DALL-E can generate a corresponding image of such a scene. This ability to generate images from text makes DALL-E a powerful tool for creative applications, such as generating illustrations, product designs, and more.
Another interesting article for the French speaking people to understand how it works.
Like other OpenAI tools, you can test Dall-E for free : https://labs.openai.com/. You can generate 50 pictures using keywords or long sentences then you have to buy credits.
Let’s play with Dall-E. Just enter the following sentence and see the result :
A blog post from vladimir. He is explaining what openAI and chatGPT are. Vladimir is so an amazing guy, what a brilliant mind ;)
output from Dall.E ⬇️
Pretty accurate isn’it ? With other AI tools designed to generate pictures, you can also enter a picture as input, your picture will be used as a baseline to generate another one. The next step will be to generate video. It’s already possible but it will be mainstream in a few months/years. As explained previously, synthetic content (music, video clips, articles) will flood on the Internet very soon. The GAFAM will soon be number one in the music chart, the best AI will be approved by the masses. I’m not so excited about this.
Lot of content creators will also use AI to generate text, pictures, videos and I’m pretty sure youtube will be full of this soon. We can also think about video games. The dev team will work to create the engine and the story, the levels, the characters design could be generated by AI and you can have a different flavor of the same game each time you play it. It seems even netflix has used AI to generate anime serie.
codex
These guys in the silicon valley think also about the concept of the digital workforce. For example, you want to build a web application, you can hire two developers and a bot. The bot has his own account and his own virtual machine and it contributes to the source code. We’re only at the beginning but an hybrid work force is a mix of human and bot, a powerful team, as explained by the salesman.
In the other way we can also ask a developer to become a super developer : a dev with his super buddy should be able to achieve more in less time. The productivity will increase and you can hire less people to do the job. We already have GitHub CoPilot powered by OpenAI Codex. I don’t think we’ll get rid of all developers but instead we’ll have super developers. Is it good or bad ? It's another question.
Let’s try to go further, if your virtual coworker or bot has its own virtual machine and its virtual account to the platform, it can also be able to deploy your source code or fix some problems. It can do the maintenance as any Ops Engineer. I think that I can try to create an IT company with virtual developers and virtual Ops Engineers using ChatGPT API, a virtual DevOps team able to code and deploy webapps to my customers… What do you think ? It’s even better than the “no code” concept. I may need virtual sales bots because I’m not good at selling services but an AI DevOps team seems interesting. I could control an army of devops bots to manage and support a big infrastructure and compete with tech giants.
I think that we will not lose our jobs but probably our jobs will evolve. As things get faster and faster, we don’t know the name and the concept of the job that our kids will do, we have no idea.
Whoever reaches a breakthrough in developing artificial intelligence will come to dominate the world.
Vladimir Poutine
Sorry for the quote but that is so true, last year I was impressed by Lamda and MuM from Google but Today openAI seems to be on top of the competition. Google is expected to announce their own chatGPT very soon, let’s see if they can create something better. Amazon seems to give up as the Alexa brand is about to be discontinued.
azure openAI
For Microsoft, the problem is different. They don’t need to compete against OpenAI, they can just buy it. They have already invested in it at the beginning ($1 Billion). I’ve heard, this year they want to bring $10 billion on the table… OK. In the meantime they have lay-off 10 000 employees. Do you see a link between these two facts ? The same trend for Google, Amazon, Twitter, Meta : fire people and invest in AI.
That does not sound very good
They say it’s not linked and that is a post pandemic effect. Why not ? It’s known that Amazon is using AI to evaluate low productive employees and fire them. In HR, a lot of things are automated, the onboarding process, check references, I guess next time I’ll search for a new job, my résumé will be checked by an AI bot and probably I’ll have to be interviewed by chatGPT or a variation of it.
This is exactly what Microsoft is trying to achieve with Azure OpenAI. You can build your own AI API i.e. your own AI tool, with your own data and your own infrastructure (at least the infrastructure powered by Microsoft Azure) instead of relying on the public chatGPT API.
So what’s next ? Who knows ? I guess that probably we even don’t know the job of our kids yet. Maybe we’ll just do some boring tasks such as labeling data for training the latest AI. I think this is already what we are doing using hashtags everywhere on social media. Maybe an AI will be able to scan everything about us from Google, Microsoft, Apple, Meta and this big AI will know everything about us and we won’t have to think anymore, just click and buy.
I’m working from home behind a computer so everything I produce can be generated by an AI so probably my job will be different in a few years. What could be my added value if everything that I think can be synthetic ? Maybe my opinion has some value in a democratic world but I'm not sure if it’s good for the business. A philosophical debate.
We still have time to use AI as a benefit for humanity, as OpenAI would like to sell it to us. We have to learn to understand it correctly and not let some big companies decide for us what will be the future of AI. Because AI will be at the top of all the decisions we will take : what to eat, what to do this weekend, where to go next summer, what to watch on TV Tonight, what job is good for me, how many kids should I have (who knows ?), even some political decisions would be made using the help of AI using financial parameters for example. For now, AI hasn't reached technological singularity yet. Let’s be prepared before it happens and use our new super buddy to achieve more and reach our goals (dreams) faster.
vladimir