[HN Gopher] Encyclopaedia Britannica Seeking $1B Valuation in IPO
       ___________________________________________________________________
        
       Encyclopaedia Britannica Seeking $1B Valuation in IPO
        
       Author : marban
       Score  : 28 points
       Date   : 2024-03-18 14:51 UTC (2 days ago)
        
 (HTM) web link (www.bloomberg.com)
 (TXT) w3m dump (www.bloomberg.com)
        
       | paxys wrote:
       | > Chicago-based firm's units include Merriam-Webster, Melingo AI
       | 
       | I'm guessing it's the last two letters that are going to get them
       | that valuation.
        
         | internetter wrote:
         | I feel like Melingo was just created for the valuation bump.
         | There's next to no information about it.
        
       | gentleman11 wrote:
       | I like the idea of an encyclopaedia company being so profitable
       | and/or valuable
        
       | wolverine876 wrote:
       | I hope this funds contracting top authors for the Encyclopaedia.
       | If you check an article's authors, you'll see that many are
       | written by experts in their field (professors, high-level
       | practicioners, etc.) and some by leading people in their field.
       | In the past they were written by people like Einstein and Freud.
       | And often articles by leading experts can be very engaging,
       | written with a personality and point of view (though they seem to
       | cover all sides).
       | 
       | But it seems like many of those authors are long retired, maybe
       | dead, and I wonder how old their contribution is (Britannica
       | seems to update it). And many articles say they are written by
       | the 'Editors of Encyclopaedia Britannica', though maybe that
       | always was the case.
       | 
       | I have nothing against the editors, but an expert in that domain
       | will have perspective, insight and current knowledge (i.e., not
       | yet in textbooks) that is impossible to match, imho. I'd love to
       | see them again hiring leading people to write.
        
         | jeremyjh wrote:
         | Luminaries like that also have their hobby horses and
         | grievances that may color their writing. The editors are people
         | doing a lifelong work of curating all kinds of information,
         | including through primary sources such as papers in peer
         | reviewed scholarly journals, such as those written by the
         | contributors you mention. So it stands to reason that there is
         | a core contributed by a leader in their field at the time,
         | which is subsequently maintained by the editors and
         | contributors.
        
           | onemoresoop wrote:
           | It would be nice to have contributions from multiple
           | luminaries on the same topic, sometimes as collaborations
        
       | justinclift wrote:
       | https://archive.is/HrwTp
        
       | justinclift wrote:
       | Hmmm, an AI that only uses factual encyclopaedia data for its
       | training set.
       | 
       | If they can completely eliminate hallucinations, they could be
       | onto something useful.
        
         | ipsum2 wrote:
         | If you train a language model on a factual dataset, it will
         | still hallucinate.
        
         | Cheer2171 wrote:
         | Do you think the reason hallucinations occur in LLMs is because
         | the training data includes the hallucination? You are very
         | wrong. Hallucinations are not a training data problem, at least
         | in the way you are implying.
        
         | barbazoo wrote:
         | Are hallucinations really only a problem with "non factual"
         | training data? Looking at all the made up library methods that
         | ChatGPT uses in the code it produces, it made me think this is
         | a general LLM/GPT problem.
        
           | rileymat2 wrote:
           | Yes, weirdly, often the hallucination is how I would have
           | liked them to expose the API.
        
         | observationist wrote:
         | Just because it's printed in a book doesn't mean it's any more
         | factual than any other source of information. Many studies and
         | reviews have found Wikipedia to be more accurate overall than
         | Encyclopedia Britannica. In fact, because it's in a book, it's
         | harder to correct, and more insidious because of opinions like
         | the one you just expressed.
         | 
         | In both cases, you should not treat the information as
         | canonically or authoritatively accurate or factual. Biases,
         | gaps, outright lies and fabrication exist in any large
         | collection of human writing.
        
           | breck wrote:
           | > Many studies and reviews have found Wikipedia to be more
           | accurate overall than Encyclopedia Britannica
           | 
           | I'm not doubting this, but do you have citations?
        
             | RandallBrown wrote:
             | I realize this isn't the most unbiased source, but it does
             | have lots of citations with studies comparing Wikipedia and
             | Britannica.
             | 
             | https://en.wikipedia.org/wiki/Reliability_of_Wikipedia
        
         | HankB99 wrote:
         | That will be interesting. I suspect that the hallucinations
         | come from combining unrelated data in ways that stretch
         | reality. However, using the general Internet as a source of
         | Truth seems like it was never a good idea.
        
         | fcsp wrote:
         | I took a brief look at their online article about the British
         | empire. While it does briefly mention Jamaica requiring
         | "conquest" in the origins section, it seems mostly oblivious to
         | the consequences of the empire's "commercial ambitions" for the
         | local populations. Not sure that would form a great "factual"
         | source of truth to train an AI on.
        
       ___________________________________________________________________
       (page generated 2024-03-20 23:00 UTC)