PulseAugur
EN
LIVE 23:29:28

Engineer builds custom AI engine, compares toddler development to LLMs

A software engineer built a custom transformer engine from scratch to understand AI models like Gemma and Llama, running them on his CPU. During this process, he observed parallels between his toddlers' language development and AI behaviors. He noted that his daughter's speech, initially incomprehensible, contrasted with the 'stochastic parrot' description of AI, while his son's sudden expression of happiness mirrored emergent properties discussed in LLMs. AI

IMPACT Provides a unique perspective on AI development by comparing it to human learning, potentially influencing how we understand AI capabilities.

RANK_REASON The item is a personal reflection and technical exploration by an individual, not a primary announcement from a major AI lab or a significant industry event.

Read on LessWrong (AI tag) →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Engineer builds custom AI engine, compares toddler development to LLMs

COVERAGE [1]

  1. LessWrong (AI tag) TIER_1 English(EN) · Carlo Valenti ·

    I ran the standard AI litmus tests on my two toddlers (yep)

    <p>In July 2022 I was in a parking lot with a Portuguese colleague, trying to fix the cargo-metering system of a 12-ton tanker truck. During a break I read a headline on my phone: <em>Google engineer claims experimental AI went sentient.</em> An engineer (like me!), from Google, …