Acuitas Diary #63: Minor progress on a variety of fronts. Narrative improvement inches forward, Text Generator refactoring is close to done, and Text Parser cleanup has begun in preparation for further feature addition.
https://writerofminds.blogspot.com/2023/08/acuitas-diary-63-august-2023.html
#ai #artificialintelligence #ChatBot #nlp #nlu
Think India: Is Sangh Parivar masquerading as academic forum in elite institutes?
The Advisory Board Committee of Think India comprises several members with direct associations to ABVP.
#radicalisation #academia #education #ThinkIndia #propaganda #misinformation #RSS #ABVP #BJP #IIT #NIT #NLU #hindutva #communalism #india
#radicalisation #academia #education #thinkindia #propaganda #misinformation #RSS #abvp #BJP #IIT #nit #nlu #hindutva #Communalism #india
After many many attempts on designing a visual #conlang that is easy to parse by software I think I'm finally on track.
Using nodes with LASERS seems to be easy and fun to build this trees! 🤩
#conlang #nlu #computationalsemantics
Acuitas' text parser gets the long-awaited support for gerunds and participles! They had to wait this long for being tricky; they're verbs, but ... not? Read further: https://writerofminds.blogspot.com/2023/07/acuitas-diary-62-july-2023.html
#ai #artificialintelligence #nlp #nlu #ChatBot #ChatBots
It's time for the final installment of Acuitas and the Symbol Grounding Problem. After all that background, I can finally talk about how I'm doing/planning to do symbol grounding in my own project! Despite lacking sensorimotor capabilities, Acuitas grounds language in his own state and function.
https://writerofminds.blogspot.com/2023/07/sgp-part-vi-acuitas-and-symbol.html
#artificialintelligence #ai #nlp #nlu #ChatBot
🧠 #PaLM2 di #Google di certo si prepara a diventare il miglior #LLM a disposizione.
🧐 Ma lasciatemi dire che, per ora, siamo lontanissimi dal livello di #GPT4.
💡 Nel test, ad esempio, ho provato a fare sintetizzare al modello un video di YouTube, facendo indicare anche minuto e URL degli argomenti.
🦾 Ho fatto diversi tentativi, ma senza ottenere output soddisfacenti. Anzi, il calcolo dei secondi è costantemente sbagliato.
#AI #IntelligenzaArtificiale #VertexAI #NLU #GoogleCloud #ChatGPT
#palm2 #google #LLM #GPT4 #ai #intelligenzaartificiale #vertexai #nlu #googlecloud #chatgpt
Having considered the objections, let's look at methods: *how* might a language-driven agent without a body do symbol grounding?
https://writerofminds.blogspot.com/2023/06/sgp-part-v-symbol-grounding-for.html
#ai #artificialintelligence #robotics #nlp #nlu #ChatBot
I'm slowly making progress on multiple areas of Acuitas. Narrative, Game Playing, and now a text generation/speech output overhaul are all moving ahead (got to give myself some variety).
https://writerofminds.blogspot.com/2023/05/acuitas-diary-60-may-2023.html
#ai #artificialintelligence #ChatBot #nlp #nlu
Ok, now the Text Meaning Representation includes the Agent and Theme. At least for this simple example "a person ates an apple". Where Ingest is the main event, with Person as the agent and Apple as the Theme.
Agent and Theme are assigned following an evaluation formula that takes into account the expected elements on the Agent and Theme field for Ingest (a Person and some kind of Food).
Here is the system creating an incomplete Text Meaning Representation for "a person ates an apple".
With more work the system should be able to add more information to the representation (agent, theme, time, modality).
Adding some tokens to ask questions, like "Do you like me?".
Also doing some work on the engine that will turn the selected tokens into Text Meaning Representations.
Right now I'm skipping any kind of Syntactic Mapping, going straight into Semantics. Constructing a tree with all possible solutions. Inspired on how Chess Engines work.
If the evaluation function is done right, the system should be able to handle ambiguity and noise.
In this month's work, I look at grounding a term that is defined relative to somebody else's actions, progress toward text adventure gaming continues, and I forgot the important detail that humans have mouths, oops:
https://writerofminds.blogspot.com/2023/04/acuitas-diary-59-april-2023.html
#ai #artificialintelligence #ChatBot #nlp #nlu
Part 3 of the Symbol Grounding Problem: is grounding even *possible* for something that's not made of real goopy neurons?
https://writerofminds.blogspot.com/2023/04/sgp-part-iii-on-brain-physics-qualia.html
#AI #ArtificialIntelligence #NLP #NLU #chatbot #robotics #ChatGPT
#ai #artificialintelligence #nlp #nlu #ChatBot #robotics #chatgpt
If you are interested on #NLU please boost this toot so we can find each other! 🔎
#nlu #ontology #semantics #knowledge
In this month's developer diary, I keep plugging away on Narrative upgrades for Big Story while also starting to introduce game playing ...
https://writerofminds.blogspot.com/2023/03/acuitas-diary-58-march-2023.html
#ai #artificialintelligence #ChatBots #nlp #nlu
🧠 Attraverso una chiamata alle API di #ChatGPT o #GPT4 verifico se un contenuto fornisce dati sufficienti per poter dare una risposta adeguata ad una richiesta.
🦾 Il JSON di output contiene la verifica, la motivazione e la risposta nel caso la verifica sia "true".
💡 Sistemi di verifica sui contenuti possono limitare il problema delle allucinazioni dei modelli di linguaggio.
#chatgpt #GPT4 #ai #intelligenzaartificiale #gpt3 #nlp #nlu
🦾 Usando le API di #ChatGPT è possibile trasformare richieste in linguaggio naturale in JSON per ricerche di prodotti su un #ecommerce.
👉 Il tutto zero-shot (senza addestramento).
Le richieste:
1️⃣ scarpe da corsa Asics numero 45 fluorescenti;
2️⃣ cerco delle scarpe per correre numero 45 della marca con il logo con le tre strisce, se possibile nere;
3️⃣ una maglietta tecnica M della Nike.
#chatgpt #ecommerce #ai #intelligenzaartificiale #LLM #gpt3 #nlp #nlu