At #Palantir, we build products that support the effective and responsible use of AI. In our latest blog series, we dive into how that commitment to #AI #Ethics and Responsible AI underpins Palantir AIP. To start, learn why the Ontology is your best defense against hallucinations: https://lnkd.in/e53RCn8R
Excellent article to help people understand the hallucinations and limitations of the glorified data chat interfaces called AI 😆. Al though there is a pun to the artificial aspect of it when you consider the hallucinations. The risk of anthropomorphizing models. “These models cannot respond, reply, or reason in the true sense of the words; rather, we provide some text as input to the model, and the model returns some statistically likely text as its output. The model is not required to be factual, when generating a statistically likely (next token) prediction based on it’s training data. What ‘looks plausible,’” otherwise known in human language as a good lie.
Love this
👍🤠
In the Business of Big Data
1moThe Ontology framework, from my experience, helps users realize, prove, and thereafter debunk so-called artificial intelligence via LLMs, which is simply the limitless permutations and combinations of more than 1.5 million English words ---- adjectives, adverbs, verbs, nouns, pronouns, prepositions, et al. --- exponentially expanded by more than 6,500 other known languages on Earth, and further exaggerated by mathematical, statistical, algebraic, calculus formulas and concepts, thereafter wrapped up in creative writing. In short, Ontology is a reminder that LLMs simply overwhelm the user with the sheer volume of information and data, because it's easier to be razmatazzed and all googly-eyed by the technology of glorified dictionaries and thesauruses rather than practice actual, real-world, fact-based, critical, intellectual, and innovative thinking. Kneel before Artificial-Intelligence-Zod!