The other side

We thought we were using large language models (LLMs) to their great potential. When I checked with my dead counterpart, they said they are using LLMs for Translation, Question Answering, Text Categorization, Program Execution, Toxic Language Detection, Textual Entailment, Information Extraction, Text Matching, Question Generation, Sentiment Analysis, Cause Effect Classification, Sentence Composition, Commonsense Classification, Text Completion, Text to Code, Answerability Classification, Data to Text, Title Generation, Gender Classification, Coreference Resolution, Paraphrasing, Fill in the blank, Named Entity Recognition, Text Quality Evaluation, Question Rewriting, Sentence Perturbation, Word Analogy, Linguistic Probing, Sentence Ordering, Dialogue Generation, Speaker Identification, Language Identification, Summarization, Dialogue Act Recognition, Fact verification, Spelling Error Detection, POS Tagging, Style Transfer, Text Simplification, Entity Relation Classification, Explanation, Speaker Relation Classification, Mathematics, Paper Revision, Entity Generation, Word Relation Classification, Negotiation Strategy Detection, Overlap Extraction, Poem Generation, Word Semantics, Spam Classification, Dialogue State Tracking, Preposition Prediction, Ethics Classification, Answer Verification, Section Classification, Story Composition, Punctuation Error Detection, Stance Detection, Stereotype Detection, Intent Identification, Code to Text, Wrong Candidate Generation, Coherence Classification, Grammar Error Correction, Number Conversion, Question Decomposition, Grammar Error Detection, Sentence Expansion, Keyword Tagging, Irony Detection, Discourse Connective Identification, Discourse Relation Classification, Sentence Compression, and Multimodal Tasks. We should also try. They are not as advanced in the LLM’s multimodal capabilities as we are. However, they have explored the textual capabilities quite exhaustively.

They use a composite model, which includes traditional AI and LLMs, to predict when someone will be born again. That is a regression model. “To be born as what?” is a multiclass classification model. They have a set of features for their model. While the features help make the predictions, the model does not understand the internal mechanics of real life (or should I say internal mechanics of real death?). This is like how the models behave for us. Nothing surprising.

Someone taking birth is like the “end of an era”. They do trend analyses of population increase and study gender distribution. They use the same time series models as ARIMA. They use the insights generated from the models to control population explosion. Prediction of birth gives them the ability to peek into the future. Our road accidents play havoc with their model’s effectiveness. This is another reason for us to follow the “Don’t LLM while driving” advice here. We can ask the deepseek-r1-32b tips to ease our traffic problems, including dangerous driving.

Let’s return to their LLMs. They do not face the challenges of ethical concerns, data sensitivity, and breaches yet. They say everything is fair in love, war, and among the dead. The extensive use of LLMs creates a new “information economy” in the afterlife, where knowledge and data become valuable commodities. This leads to power dynamics and social structures based on AI expertise. The dead counterparts are close to discovering a way to use their advanced LLMs to communicate with the living, bridging the gap between worlds. This might lead to ethical dilemmas and potential disruptions in both realms. Talk about the risks associated with interdimensional communication! As LLMs answer age-old questions about life, death, and the universe, the dead grapple with the loss of wonder and mystery.

Does true advancement lie in pushing the technological boundaries of LLMs or in accepting the limits of our understanding? Some aspects of existence remain fundamentally unknowable and must stay like that.

Existential questions are screaming at them. As the dead push the boundaries of LLM capabilities, they begin to question the nature of their existence and the purpose of their continued “life” after death. While concluding this article, I realised that my counterpart among the dead is writing about us and how we use LLMs in our lives and getting it published!