AI Safety seminář

[Log in here.] (Https://Airtable.com/shrmtl0n5hxxvhrca?utm_source=ais-seminar)

The deadline for applications is 3.3.2025. The eight -week seminar starts in the week from 10.3.2025.

** Artificial intelligence is a huge opportunity, but also one of the greatest risks for the future of humanity. How to insure that future transformative artificial intelligence systems turn the world significantly for the better, not vice versa? Why is AI considered an existential risk and significant experts in the field of artificial intelligence research? **

General artificial intelligence (AGI) may not have consciousness as a person, yet in the relatively close future, it can out renowned a person in many ways. At present, we do not know how to create such AI so that we can guarantee that it is not a threat to humanity.

We have prepared an eight-week seminar according to [program] (https://www.eacBridge.org/agi-safety-fundamentals), which originally originated at the University of Cambridge. You can log in to one of the seminar versions. One seminar is more [technical] (https://course.aisafetyfundamentals.com/alignment) and the other is focused on [AI Governance] (https://course.aisafetyfundamentals.com/governance), ie rectification via standards, regulations and foreign policy approaches to international security.

** For whom is the seminar intended? **
You can apply for both versions of the seminar without previous knowledge of Machine Learning, if you are willing to devote additional time to materials explaining the basic principles of machine learning in the first week.

Sign in to understand why current AI development methods do not have its long -term safety for society and humanity, and you want to learn what to do with it.

The seminar can be followed by practical work projects concerning eg evaluation of large language models.

** 💡 What will you learn? **
** Technical Security Seminar AI **

* Brief intro into neural networks and how AI trains
* What is general artificial intelligence (AGI) and how quickly we are heading to it
* How AI can create instrumental goals, ie why it could resist the shutdown or why it could start to accumulate resources.
* How missed a reward function can lead to the optimization of undesirable behavior (Reward Hacking, Reward Misspecification)
* How can AI erril to generalize their goals in training
* How to scalance the possibility of supervision of AI if we stop understanding the problems it solves
* What are the currently proposed procedures for safe training AI
* What does research look like interpretability, that is, the ability to understand what AI “thinks”
* How AI represents agents and agency within

The whole syllabus can be found [here] (https://course.aisafetyfundamentals.com/alignment).

** Seminar O AI Governance **

* Brief intro into neural networks and how AI trains
* What specific catastrophic risks of poorly set AI face
* Why the safety of AI is difficult to reach
* What standards and regulations can help with the safe development of AI
* How to limit the possibilities of developing dangerous AI systems to dangerous actors
* How to drive ai risk through regulation of chips production
* How to ensure international cooperation

The whole syllabus can be found [here] (https://course.aisafetyfundamentals.com/governance).

** 🎓 How does the seminar go? **
You will read texts from different sources in English every week. Preparation takes about 3 hours a week or more (depending on your care).
You will be divided into smaller groups of 3 to 6 people who will meet once a week for an hour to an hour and a half long discussion, where you will discuss the read texts for the week. The discussion itself takes place in Czech.

The discussion will help you understand and organize your thoughts and also get to know interesting people who are interested in AI and its safety. The discussion will be moderated by an experienced facilitator that helps to care for its quality and ensures a safe and stimulating environment.

** ⏳ When? The meeting will be 8 or 9 - one week one.

** 🗺️ Where? The deadline for applications is on Monday, March 3. In the week after the deadline, specific regular meetings will be arranged according to the time possibilities of the participants in the individual groups. The meeting will be 8 - one week one.

** 🏷️ Price: ** FREE

👉 **

** Don't have time? **
If you are interested in AI, but for time reasons you cannot participate in the seminar, let us contact yourself and we will contact you with other opportunities to engage (events, discussions, lectures, projects).

[I want to be contacted/a] (https://forms.gle/jqvafwphpzmiffy8)

Don't you want other similar opportunities to escape? You can find out about the news at our [website] (https://efectivni-altruism.cz/) or on the [Facebook pages] (https://www.facebook.com/efectivnialtruism), or you can log in. Newsletter.