Ευχαριστούμε που στάλθηκε η αποσαφήνισή σας! Ένα μέλος της ομάδου μας θα επικοινωνήσει μαζί σας σύντομα.
Ευχαριστούμε για την εκδήλωση κράτησης! Ένας από τους συνεργάτες μας θα επικοινωνήσει μαζί σας σύντομα.
Διάρκεια 7 Ώρες
Εξέλιξη Κομματιού
- Day 1: Handing over tasks in the browser
- How browser agents work
- Reading a page, planning, clicking, typing and checking
- Difference from research in a chat
- Where they are reliable today and where they are not
- Exercise: Observe an agent on a simple task and note down every step.
- The tools compared
- ChatGPT Agent and the Atlas browser
- Claude in Chrome as an extension
- Perplexity Comet
- Availability, cost and data protection
- Exercise: Solve the same research task with two tools and compare them.
- Research and comparisons
- Comparing vendors, prices and terms across several pages
- Results as a table with links
- Cross-checking on the original pages
- Exercise: Have a vendor comparison with five criteria created and verify two values.
- Forms and web applications
- Filling in forms with existing data
- Transferring data from web applications into tables
- Stopping before submitting, paying or deleting
- Exercise: Have a test form filled in with data from a table and review it before submitting.
- Security
- Hidden instructions on web pages (prompt injection)
- Logged-in accounts, passwords and payment data
- Separate browser profiles and approvals
- Exercise: Open a prepared test page with a hidden instruction and evaluate the agent's behavior.
- Rules and usage plan
- Which tasks are allowed
- Who reviews and approves
- Planning recurring tasks
- Exercise: Write a page of rules and three use cases for the own team.
- Outcome: Each participant has handed over two tasks to a browser agent, verified the results and defined rules for use in the team.
- How browser agents work
Απαιτήσεις
- Laptop with Chrome, an account for at least one of the tools (ChatGPT, Claude or Perplexity), and a test account without sensitive data for the exercises.
- No programming skills required.
Audience
- Specialist users, assistants and researchers who want to hand over recurring work on web pages and in web applications to AI agents.
Σχόλια (3)
Συνοπτική επισκόπηση του υποβάθρου· τρόπος σύνταξης μιας εντολής (prompt).
Olaf Kannt - Helios Klinikum Krefeld
Κομμάτι - Generative AI and Prompt Engineering in Healthcare
Μηχανική Μετάφραση
οι συμβουλές και οι προτεινόμενες εντολές που μπορούμε να αντλήσουμε από αυτή την εκπαίδευση
Lee Mei Lin - ST Engineering IHQ Pte Ltd
Κομμάτι - InVideo AI: Creating Engaging Short-Form Videos
Μηχανική Μετάφραση
χρήση κατάλληλων και αποτελεσματικών προτροπών
Marses Pacaldo
Κομμάτι - Generative AI and Prompt Engineering for Corporate Professionals
Μηχανική Μετάφραση
