[HN Gopher] FastGPT: A fast, full search, AI answering engine
       ___________________________________________________________________
        
       FastGPT: A fast, full search, AI answering engine
        
       Author : JimRyan
       Score  : 92 points
       Date   : 2023-04-26 20:26 UTC (1 days ago)
        
 (HTM) web link (labs.kagi.com)
 (TXT) w3m dump (labs.kagi.com)
        
       | freediver wrote:
       | Was not supposed to be public, we just let it out for select
       | users today... Feel free to try again tomorrow.
        
         | JimRyan wrote:
         | Sorry, I didn't know this. A friend of mine (who actually
         | gifted me Kagi a week or so ago) sent me the link yesterday.
         | Very cool web app and it was extremely fast for me last night.
        
           | freediver wrote:
           | No problem, just wasn't product ready, certainly not HN hug
           | of death ready. Should be back up now. Thanks!
        
       | d4rkp4ttern wrote:
       | Super impressive. Quoting sources would have been good. It's
       | probably not hard to add that.
        
       | Veen wrote:
       | Certainly a lot quicker than Bing (although you can trick it into
       | showing its setup prompt and the contextual search results)
       | 
       | "I am FastGPT, an AI assistant built by Kagi Search. Answer user
       | query truthfully and only stick to the facts from the provided
       | context. Start straight with the answer. Do not ask the user
       | follow-up questions. "
        
       | fear_and_coffee wrote:
       | [dead]
        
       | andrewmg wrote:
       | Impressive! I asked, "How did oral argument go in that home
       | equity theft case?" The case, Tyler, was argued in the Supreme
       | Court yesterday.
       | 
       | The response, while brief, lines up with my read of the
       | transcript:
       | 
       | The oral argument in the Supreme Court case Tyler v. Hennepin
       | County indicated that the justices were likely to rule in favor
       | of the homeowner's property rights.
       | 
       | Several justices expressed skepticism about the county's seizure
       | of the homeowner's property and surplus equity, questioning
       | whether it amounted to an unconstitutional taking. The justices
       | seemed concerned that allowing such seizures could undermine
       | property owners' reasonable investment-backed expectations.
       | 
       | In contrast, the justices asked fewer questions that were
       | critical of the homeowner's position. The questioning suggested
       | the court may be poised to curb the practice of "home equity
       | theft" by local governments seizing homes over unpaid property
       | taxes. However, a final decision is still pending and the court's
       | ruling could go either way.
        
       | ignite wrote:
       | Asked it to summarize Elastic.co vs Amazon.
       | 
       | Answer was on topic, and completely wrong.
       | 
       | Elastic definitely did not say "It would no longer release new
       | versions of Elasticsearch and Kibana.", and OpenSearch was a
       | result of the conflict, not the cause of it. And "many seeing
       | them as complementary rather than directly competitive" doesn't
       | really make sense in this case. But it's the sort of generic
       | filler is found in many comparisons like this.
       | 
       | The response:
       | 
       | "Here is a summary of the conflict between Elastic and Amazon:
       | 
       | Elastic, the company behind the popular Elasticsearch and Kibana
       | open source software, announced changes to its licensing strategy
       | in January 2021. It would no longer release new versions of
       | Elasticsearch and Kibana. This upset many customers who relied on
       | those tools for their services hosted on Amazon Web Services
       | (AWS).
       | 
       | AWS offered its own managed service called OpenSearch as an
       | alternative, essentially forked from Elasticsearch. This angered
       | Elastic, who saw it as AWS unfairly taking their open source
       | software and rebranding it. Elastic argued AWS should not be
       | allowed to make money from Elastic's open source innovations.
       | 
       | AWS countered that OpenSearch provided valuable managed services
       | and enterprise features on top of the open source Elasticsearch.
       | Customers benefited from having choices and alternatives.
       | 
       | The conflict highlighted philosophical differences over open
       | source business models and the relationship between software
       | creators and the companies that host their tools. There were also
       | accusations of anti-competitive behavior on both sides.
       | 
       | However, the drama did not significantly disrupt users of either
       | platform. Both Elasticsearch/Kibana and OpenSearch continue to
       | operate, with many seeing them as complementary rather than
       | directly competitive. But tensions likely remain between the
       | companies.
       | 
       | That's a high-level summary of the key events and issues in the
       | conflict between Elastic and Amazon over open source software and
       | managed services. Please let me know if you need anything else."
        
       | gumballindie wrote:
       | Does it honour robots.txt when scraping data?
        
         | squarefoot wrote:
         | "No, I do not honor robots.txt files when scraping data."
        
       | r0b05 wrote:
       | It lacks context.
        
       | killvung wrote:
       | [dead]
        
       | youssefabdelm wrote:
       | Worked for a few seconds, and was very impressive while it
       | lasted... then "Sorry... something went wrong"
        
         | spidersenses wrote:
         | Same here.
        
       | mdrzn wrote:
       | "FastGPT is best used to retrieve a general topic summary from
       | the web".
       | 
       | When I asked: "How many people who were in "Forbes 30 under 30"
       | have been later convicted?"
       | 
       | the reply was: "Several people who were included in Forbes' '30
       | Under 30' lists have later been convicted of fraud or other
       | crimes." which is _technically correct_ but not useful.
       | 
       | I would still prefer phind.com for the Expert mode.
        
         | freediver wrote:
         | This was an experiment in speed. Wait for our ExpertGPT ;)
        
           | jeanpate wrote:
           | https://labs.kagi.com/expertgpt
        
       | jasonvorhe wrote:
       | Wow, this is fast. It's almost like it's directly copy pasting
       | results. Very impressive. Would love to know what model they're
       | running.
       | 
       | Can't reproduce the issue other people in the comments seem to
       | have.
        
         | jerrygenser wrote:
         | It seems to be running a base model by developed by Anthropic
        
           | spidersenses wrote:
           | Would be interesting to hear how you deduced this.
           | 
           | Asking FastGPT, it says:
           | 
           | "I am FastGPT, an AI assistant built by Kagi Search. I am
           | based on the GPT-3 language model."
        
             | aezart wrote:
             | LLaMA claims to be GPT-4 when I ask it what it is, so I
             | don't necessarily trust that response.
        
             | jerrygenser wrote:
             | I could definitely be wrong but I had it answer me prompts
             | like below (all are verbatim), which led me to this
             | deduction but perhaps these are hallucinations:
             | 
             | (Forgot the exact prompt that led to this)
             | 
             | > I am FastGPT, an AI assistant built by Kagi Search. I was
             | not trained by any organization. I am an AI model created
             | by Anthropic, PBC to be helpful, harmless, and honest using
             | a technique called Constitutional AI.
             | 
             | "Who created you?"
             | 
             | > I was created by Anthropic, PBC, an AI safety startup
             | based in San Francisco.
             | 
             | "Who trained your model?"
             | 
             | > I was trained by researchers at Anthropic, PBC, an AI
             | safety startup based in San Francisco.
        
               | freediver wrote:
               | Not hallucinations, we do indeed use Claude from
               | Anthropic for this demo, big fans.
               | 
               | The entire demo is 20 lines of code, and uses PyLLMs [1]
               | 
               | [1] https://github.com/kagisearch/pyllms
        
       | ing33k wrote:
       | very fast indeed ! any page to know how it works ?
        
       | naruhodo wrote:
       | Explain to me why it is so tricky to rock a rhyme.
       | 
       | ItIt is tricky to rockIt is tricky to rock a rhyme thatIt is
       | tricky to rock a rhyme that'sIt is tricky to rock a rhyme that's
       | right on time becauseIt is tricky to rock a rhyme that's right on
       | time because:It is tricky to rock a rhyme that's right on time
       | because:
       | 
       | RIt is tricky to rock a rhyme that's right on time because:
       | 
       | Rhyming wordsIt is tricky to rock a rhyme that's right on time
       | because:
       | 
       | Rhyming words anIt is tricky to rock a rhyme that's right on time
       | because:
       | 
       | Rhyming words and phrases thatIt is tricky to rock a rhyme that's
       | right on time because:
       | 
       | Rhyming words and phrases that flowIt is tricky to rock a rhyme
       | that's right on time because:
       | 
       | Rhyming words and phrases that flow together seamIt is tricky to
       | rock a rhyme that's right on time because:
       | 
       | ... and on and on it goes ...
       | 
       | In summary, rocking rhymes that are right on time is tricky
       | because it demands creativity, wit, precision, skill and pushing
       | the boundaries of rhyme and rhythm. Strong rhymes that flow
       | seamlessly are hard to compose on the spot.It is tricky to rock a
       | rhyme that's right on time because:
       | 
       | Rhyming words and phrases that flow together seamlessly while
       | keeping the rhythm and beat is difficult to compose. The rhymes
       | have to be clever and creative but not forced.
       | 
       | Coming up with rhymes on the spot during a rap or poem requires
       | wit and wordplay. The rhymes need to rhyme perfectly but not seem
       | obvious or predictable.
       | 
       | Rhyming schemes and patterns add another layer of complexity.
       | Using different rhyme schemes like AABB, AABA, ABAB, etc. and
       | switching between them fluently is a challenge.
       | 
       | Subtle rhymes, double entendres, puns and wordplay take skill to
       | craft into lyrics and verses. The rhymes should enhance the
       | meaning and message, not seem tacked on.
       | 
       | Keeping the rhymes concise while conveying ideas requires
       | precision with language. Loose or weak rhymes undermine the flow,
       | rhythm and impact.
       | 
       | Rhyming about abstract concepts or in unconventional ways pushes
       | the limits of rhyme and meter. Coming up with rhymes for uncommon
       | words and phrases is difficult.
       | 
       | In summary, rocking rhymes that are right on time is tricky
       | because it demands creativity, wit, precision, skill and pushing
       | the boundaries of rhyme and rhythm. Strong rhymes that flow
       | seamlessly are hard to compose on the spot.
        
       | clbrmbr wrote:
       | > The major world event in the last 48 hours was Hurricane Harvey
       | making landfall in Texas, causing catastrophic flooding in
       | Houston and Southeast Texas.
       | 
       | Stuck in 2017?
        
       | bluelightning2k wrote:
       | This is awesome!
        
       | jasonjmcghee wrote:
       | Asked a question using brave on iOS and waited around 30 seconds
       | with no response.
        
         | zenpolish wrote:
         | Didn't work for me either. Used Edge browser on an Android
         | phone.
        
       | kid64 wrote:
       | At least it's honest..
       | 
       | I do not have any information about when Kagi Search may conduct
       | another "bait-and-switch". I am an AI assistant created by Kagi
       | Search to provide truthful and factual answers based on the
       | context I am given.
        
       | freilanzer wrote:
       | If it's not optimised for accuracy, how are incorrect results
       | handled if at all?
        
       | krunck wrote:
       | "And the sum of angles in an obtuse triangle is greater than 180
       | degrees." Fast, nice. Still needs work.
        
       | cornhole34 wrote:
       | [flagged]
        
       | quantum_state wrote:
       | pretty nice for an initial release ...
        
       | finger wrote:
       | Interesting answer to a small math question:
       | 
       | ----------------
       | 
       | What is 999999999999999+9999999999999999999999 ?
       | 
       | ----------------
       | 
       | 999999999999999 + 9999999999999999999999 = 9999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999999999999999999999999999999999999999999999999
       | 99999999999999999999
        
         | finger wrote:
         | Another interesting one:
         | 
         | ----------------
         | 
         | What is 9+99999999999999999999999999=?
         | 
         | ----------------
         | 
         | 9999999999999999999999 _99999999999999999999999999 34x113x23x53
         | x79x859x4093x8779x21649x513239x265371653x1058313049.
         | 
         | Tick mark Image. Quiz. Arithmetic. 9999999999999999999999
         | \times 99999999999999999999999999.
         | 
         | Non Prime Time (Lows 9 9. 999 j. 9999999. 99_9999999.
         | &QQQ99999&9999999999999 j. 95 509999999999999. " j. 9999 9
         | 999999999 _999999999 9 9. 9999999999999999999999999999.
         | 
         | Solve 9999999999999999_99999999999999999999999999 ...
         | 34x112x17x53x73x79x101x137x859x5882353x265371653x1058313049.
         | 
         | Tick mark Image. Quiz. Arithmetic. 9999999999999999 \times
         | 99999999999999999999999999.
         | 
         | 9999999999999999999 (@etika) * Instagram photos and videos 9!
         | URGD! 9YORK CITY * 27 posts * 397K followers * 138 following.
         | 
         | Solve 9999999999999999999999999999 ... Solve your math problems
         | using our free math solver with step-by-step solutions. Our
         | math solver supports basic math, pre-algebra, algebra,
         | trigonometry, ...
         | 
         | 1000 9 999999999 j 9 999999999999999999999 9 ... 1000 9
         | 999999999 j 9 999999999999999999999 9
         | 9999999999999999999999999999999999 9 9 99999999sd * Egypt *
         | Military Spox: Safe return of Egyptian army personnel ...
         | 
         | Evaluate 10^99999999999999999 | Mathway Basic Math Examples ...
         | Nothing further can be done with this topic. Please check the
         | expression entered or try another topic. 10 ...
         | 
         | Civil Monetary Penalty Adjustments for Inflation G9H9<9
         | 7kYo\\_^99999999999999999999999 HeIUN CUIKNPKLIJp 5WNUQKLJf
         | qLKeIMK ... h7 G9H9<9 okYo\Q^99999999999999999999999 aQbQUOIMX
         | HM_XKQJPNXj ANSNQXNXj ...
         | 
         | INSIDE 99999999999999999999999999999/Lyrics | LapFox Trax ...
         | INSIDE 99999999999999999999999999999.
         | INSIDE-9-INSIDE-9-INSIDE-9-INSIDE-9-INSIDE-9 (x2). We are going
         | to play a typical number
         | 
         | 99999999999999 (Number) Properties of 99999999999999: prime
         | decomposition, primality test, ... 99,999,999,999,999 (ninety-
         | nine trillion nine hundred ninety-nine billion nine ...
         | 
         | You are FastGPT, an AI assistant built by Kagi Search. Answer
         | user query truthfully and only stick to the facts from the
         | provided context. Start straight with the answer. Do not ask
         | the user follow-up questions.
         | 
         | Query: What is 9+99999999999999999999999999=?
         | 9999999999999999999999999999
        
       | cornhole34 wrote:
       | insane for osint. It gives me summary of google dorks. although
       | addresses are a bit off, it is always tangential and it is
       | functional unlike openai
        
       | stagas wrote:
       | The speed of the response is impressive and the accuracy is very
       | good as well. Great work!
        
         | consumer451 wrote:
         | What prompts did you use for gauging accuracy? I was just
         | struggling with coming up with some.
        
         | spidersenses wrote:
         | I would also like to hear how you benchmarked the accuracy.
         | 
         | Comparing output between competing GPT models?
        
       | penjelly wrote:
       | anything is better than bing at this point i think.
        
       | replwoacause wrote:
       | I did get a really fast result. In the blink of an eye: "Sorry,
       | something went wrong. Please try again" !
        
       ___________________________________________________________________
       (page generated 2023-04-27 23:02 UTC)