- AInauten.net
- Posts
- π₯ Weekly AI news: Did you miss it?!
π₯ Weekly AI news: Did you miss it?!
π¨βπ The most important AI updates at a glance
Die AInauten
May 18, 2024
This issue is brought to you by:
π₯ Weekly AI news: Did you miss it?!
π¨βπ The most important AI updates at a glance
AI-HOI, AInauts!
Maybe you didn't catch all the news, tools, and hacks about AI last week, or maybe you've only recently joined us. Either way, here's our recap with all the headlines from the newsletter - just one click away!
Click the links to jump right to the article - or read our picks below.
β Selection of the top posts of the last week β
π¦ΈββοΈ AI + automation = how to use your superpower
Last week's topics with building your own app and a phone agent have been very well received by you! Today, we would like to encourage you to take 10 minutes and try something out for yourself.
This time we are not using Zapier, but Make.com. Make offers similar features but is slightly cheaper and also offers helpful scenario-templates.
The topic: We are building a mini automation with which you will automatically generate reply suggestions for certain emails!
Email management is still an important but often time-consuming activity for most of us - but we can change that. Here's how!
All you need is a free Make account, a mailbox and an OpenAI API key.
Then click here and use the scenario template from Make. Simply connect your account for each of the steps - you can make adjustments later.
Then activate the scenario and drag an email into your defined folder.
After a few minutes, a reply will appear magically in your drafts, and it will look like this:
We consider the ability to combine AI, automation and no-code tools to be one of the most important skills for the coming years.
Thanks to AI tools that are accessible to everyone, we all have the opportunity to build impressive things. Even as a complete non-technical person. And this was just a tiny taste of what's possible!
If there is enough interest, we will be holding a free live workshop on the topic of automation with AI.
Sounds like something youβd like to attend?
Then just reply to this email with βIβm in!β, and weβll get back to you with the details once weβre ready.
𧬠AlphaFold 3 - GPT-5 of biomedicine?
All right, now let's take a look into the future: what if everything gets better, thanks to AI? We don't want to bore you with abstract facts, but instead have put together a few examples that will become reality thanks to Google's new AlphaFold 3. The avalanche of discoveries is on its way ...
via ChatGPT DALL-E 3
Without technical jargon, AlphaFold will make many things possible ... ‡
Better diagnostics: New, cheaper screening tests for cancer and other diseases can be developed.
Deeper understanding of disease mechanisms: Insights into the molecular basis of diseases such as cancer or Alzheimer's.
Precise medication & vaccinations: Personalized therapies and treatment of genetic disorders, including for rare diseases.
Higher resistance: Modification of plants so that they can withstand drought, cold, pests or diseases.
Better crop yields: Breeding plants that use carbon dioxide more efficiently and therefore produce higher yields.
Improved preservation: Proteins that inhibit bacterial growth could be used for longer-lasting products.
Remedy environmental damage: New proteins can break down pollutants, decompose plastic or neutralize microplastics in water.
Carbon capture and storage: Design of proteins that efficiently capture and store CO2.
More efficient production of bioethanol: Improvement of enzymes that convert biomass into energy.
Pretty cool, isn't it?
It almost seems as if every new AlphaFold version of Google DeepMind pushes the boundaries of the possible further towards tangible reality. A more in-depth video can be found at @2minutepapers, and Demis Hassabis also finds the right words.
Last, but not least: The AlphaFold server allows researchers to use it for free (1.8 million are already active)! You can try it out here - and let us know what your use case is.
P.S.: AlphaFold is not alone, there are many new achievements in science!
The New York Times explores the world of GenAI and CRISPR genetics in this great article.
OpenCRISPR provides free open source tools for researchers.
TIME magazine asks: "Are we ready for the ChatGPT moment in biotech?"
Google recently introduced Med-Gemini, medically trained models for diagnosis and clinical questions.
After almost 60 years, AI has finally led to the discovery of another antibiotic.
2/3 of the top pharmaceutical companies prohibit their employees from using ChatGPT.
Moderna, on the other hand, allows to create custom GPTs!
πͺπΊ Europe's biggest problem for the (AI) future
A super exciting panel discussed the AI situation in Europe. As you may know, Europe is not necessarily characterized by large AI companies.
Instead, the EU has decided it wants to be a leader in another area: regulation π
And the EU AI Act has achieved this.
In fact, the thinking was as follows (if we have understood the EU policymaker correctly):
Nobody cares about regulation yet.
We take an extremely professional approach.
At some point, other countries will also deal with regulation.
And then they will adopt a lot of things from us, because they have already been tried and tested.
Our EU AI companies will then have an advantage because they are already implementing all of this.
Don't get us wrong. We have nothing against regulation, we even think it's important. But we should set our priorities elsewhere first.
The positive thing is that this is slowly taking hold at both EU and German level. The focus is moving away from regulation and towards application.
Ideas for this include, after the Green Deal, also pushing for a tech/AI deal at European level and thus merging two large areas.
The biggest problem in Europe is ...
But what is the main problem that we seem to have in Europe?
We have excellent research and technology in terms of AI, be it the Fraunhofer Institute, UnternehmerTUM - Center for Innovation & Business Creation etc. - extremely important work is being done here.
There is still too little capital and willingness to invest, but this is also changing, as can be seen from the example of the Schwarz Group.
According to one panel member, however, Europe has a different problem:
Our quality of life is simply so high that we often lack the hunger for success! We have not enough people who want to become entrepreneurs. We need more people who are able to turn technology into a sustainable business model and sell services and products successfully.
The big opportunity (also for non-technical people)
This was one of the biggest personal learnings from the conference. Especially because we are not hardcore tech heads.
We have great technology here in Europe and also more and more financial capital. But what we need are courageous generalists who have an understanding of products and marketing - and can understand and apply what is possible with AI.
If you are such a person, find technology partners, develop products with them and market them!
Letβs MEGA!
(Make Europe Great Again π)
π₯ The most important AI trends & developments for 2024
In this section, we would like to summarize a few very cool insights about current trends and the status quo.
They come predominantly from a presentation by the creators of the AI Index Report at Stanford University. It really is an ingenious report that measures and analyzes the trends in AI every year and draws conclusions from them.
You can read the full 300+ pages of the report here for free.
Here are a few important learnings:
AI is improving at an incredible speed
Above, you can see the result of a Midjourney prompt to create a Harry Potter image. On the left with the first model in February 2022 and then again in December 2023.
Crazy.
That's why we always recommend that you donβt invest too much into developing something yourself, and ideally build your AI applications in such a way that the underlying foundation model can be changed. Depending on which one is improving faster, you are very flexible this way.
It's crazy how much the new models cost to develop
The cost estimate above shows that the playing field for these foundation models will only be dominated by large players with virtually unlimited money.
The first Transformer Large Language Model had training costs of less than 1 million euros.
GPT-4 is already just under 80 million. And the current Google Gemini Ultra is almost 200 million US dollars. And that's just the training costs!
It will be very difficult for providers with little capital to keep up in the long term.
AI is already making us much more productive
As the saying goes: It's not AI that will replace us immediately, but people who use AI will replace those who don't.
This is also visualized by the results of a study among management consultants.
Accordingly, the use of GPT-4 alone increases productivity by 12%, speed by 25% and the quality of work by 40%.
Extremely impressive figures, which are perhaps another motivation to learn AI and figure out how to use it best. Others are already doing it very successfully!
The following chart impressively illustrates the impact of this at company level:
In summary, it shows: AI is already reducing costs and increasing sales in almost every area.
You don't need a business degree to see that this is a good combination. π
Everyone gets their personal employee, teacher, consultant, β¦
We have saved the best trend for last.
OpenAI has just introduced their newest GPT-4o model. A model that is multimodal. In other words: it can speak, hear, see, read and understand!
This means that in the very near future, we will always have our smart little helper with us to see what we are doing, to talk to and to help us directly.
Watch the video here:
In general, AI will fundamentally change the way we interact with our phones and computers.
It's really worth taking a look at the entire GPT-4o product launch:
Or the one from Google. Here too, some super exciting innovations have just been announced. Check out the most important ones in this summary in under 10 minutes:
We will delve go into everything in more detail in the near future and look at individual cases.
That's enough for today. See you soon with a fresh round of news, hacks and insights!
Your AInauts, Fabian & Reto
Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.
π Please rate this issue:Your feedback is our rocket fuel - to the moon and beyond! |