Posts

In one recent run I had local LLMs going for more than 28 hours, chewing through 437 markdown files. The model used was qwen3.5:122b, served via Ollama working two Mac Studios on a private LAN. The task was unglamorous: read a file, return a JSON with a two-sentence summary, the named persons mentioned, the topic tags, the key-ideas the file argues. Repeat. Aggregate across files. Write per-entry scaffolds.

The doctoral school of the University of Lleida has invited me to give a conference (keynote) at their PhD day on 6/3/2024. Here you will find the audio and video of the conference:

ChatGPT gives me a summary of the article in The Guardian: https://www.theguardian.com/commentisfree/2024/feb/13/software-student-cheated-combat-ai Robert Topinka, a lecturer at Birkbeck, University of London, explores the dilemma professors face with students’ use of AI to write essays. After flagging an essay as “100% AI-generated,” Topinka lands in a tough spot when an exceptionally bright student challenges the accusation. The case highlights the problems with AI detectors like Turnitin, which can mistake legitimate technological support used by students for cheating. Topinka argues we must adapt academic assessment to the AI era, proposing alternatives such as presentations and podcasts to demonstrate students’ critical and original thinking, avoiding unfair accusations and promoting equal educational opportunities.

ChatGPT summarizes the article in The Guardian for me: https://www.theguardian.com/commentisfree/2024/feb/13/software-student-cheated-combat-ai Robert Topinka, professor at Birkbeck, University of London, explores the dilemma professors face with students using AI to write essays. After detecting an essay marked as “100% AI-generated”, Topinka finds himself in a difficult situation when an exceptionally brilliant student challenges this accusation. The case highlights the challenges of AI detectors, like Turnitin, which can confuse students’ legitimate use of technological support with cheating. Topinka argues for the need to adapt academic assessment to the AI era, proposing alternatives like presentations and podcasts to demonstrate students’ critical and original thinking, while avoiding unfair accusations and promoting educational equality.

In this post, I present an important document prepared by the UPC Doctoral School. It is the study “Data and applications: analysis of habits in data management and application use among doctoral students at the UPC North Campus”, based on a survey and a focus group carried out during the months of June and July.

%%{init: { ’logLevel’: ‘debug’, ’theme’: ‘default’ , ’themeVariables’: { ‘cScale0’: ‘#FFD700’, ‘cScaleLabel0’: ‘#000000’, ‘cScale1’: ‘#228B22’, ‘cScaleLabel1’: ‘#FFFFFF’, ‘cScale2’: ‘#00008B’, ‘cScaleLabel2’: ‘#FFFFFF’, } } }%% timeline title Modern Education Timeline section XIX Century 1800s : Compulsory Education Laws : Schools and Classrooms reflect a factory-like structure : Goal to alphabetize the whole society : Numerical Grading introduced by William Farish : Standardized Testing emerges : Courses by mail section Early section XX Century Early 1900a : Modern School Movement begins : Dewey emphasizes experiential learning : Montessori promotes child individuality : Pressey Teaching Machine (1926) 1940s and 50s : Piaget introduces developmental stage theory : B.F. Skinner develops behaviorism : B.F. Skinner Bulilds GLIDER teaching machine : Radio Broadcasting of Lectures : IBM teaching software section Late XX Century 1960s : First applications of computers in education : Seymour Papert, LOGO Programming Language : Computer-Aided Instruction (CAI) : IBM 1500 Instructional System 1970s : PLATO Instructional System : Distance Universities (OU , UNED, … ) : VCRs : Microcomputers , BASIC : Using computers become a subject of study

Table Of Contents # AI-Generated Content Search Engine or Database Content Dealing with Hallucinations What Are Hallucinations? Hallucination Mitigation Strategies Practical strategies and use cases of Chatbots like ChatGTP Language Considerations: for best results, use English Putting Context i context. Text Formatting: a neat trick in LLM Conversations Use text formats like a ninja Other uses of text formats and chatbots. Large Language Models (LLMs) like ChatGPT are a new family of technologies that offer a lot of potential but also come with their own set of challenges. We’re using these AI tools more and more for everything from simple questions to generating complex code. So, it’s important to ask: Do we really understand how to get the best out of them?

Table Of Contents # 1 Embeddings 1.2 Practical Applications of Embeddings 2 The context in LLMs based on transformers** 2.1 Context as Ad Hoc Training 3 Self-Referential Context and Programmability in LLMs**_ References Generative AI models, especially Large Language Models (LLMs) like GPT and its successors, have become a pivotal force in the advancement of artificial intelligence. While these models have gained prominence for their capabilities in natural language processing (NLP)—including tasks such as sentiment analysis, machine translation, and content generation—their applications extend beyond the realm of NLP [11].

Oceans as tools for capturing carbon Oceans are a massive carbon sponge, absorbing a quarter of global CO2 emissions. That makes them a powerful tool to fight climate change. A company called Equatic is testing a bold idea to remove carbon dioxide from the atmosphere: pull it directly out of the ocean. The company is piloting this idea with a barge in Los Angeles that removes 100 kg of CO2 from seawater every day.

Organizations linked to free software such as Creative Commons, Github, Huggingface have made public a document in which they offer suggestions on how to improve the AI Act to protect and foster the development of free software.

The Enigma machine, the cryptographic contraption that played a key role during the Second World War, is a powerful symbol of our technological past. Developed by Nazi Germany to encrypt its messages, this machine was considered inscrutable until the arrival of British mathematician Alan Turing, who managed to decipher its code in 1943.

Found that on the former Twitter and I thought is was on point.