Post ATtpg9quUHi4wl5kBM by phildini@wandering.shop
(DIR) More posts by phildini@wandering.shop
(DIR) Post #ATthaM0Pad6XGiG1Z2 by simon@fedi.simonwillison.net
2023-03-23T05:29:16Z
1 likes, 0 repeats
I've been puzzled at why Google Bard seems to be much less capable than Bing - for example it seems much more prone to hallucination when it should be using facts it pulls in from Google searchTurns out it's built on Google's cheaper, less powerful model!> We’re releasing it initially with our lightweight model version of LaMDA. This much smaller model requires significantly less computing power, enabling us to scale to more users, allowing for more feedbackhttps://blog.google/technology/ai/bard-google-ai-search-updates/
(DIR) Post #ATthqtx19M1LBIYnia by simon@fedi.simonwillison.net
2023-03-23T05:32:16Z
0 likes, 0 repeats
I get their desire to keep costs down and onboard more users faster... but it's VERY clear to anyone who's comparing these models that Bard is significantly behind ChatGPT 3.5, ChatGPT 4 and BingHere's a solid comparison of the big five current public models that echoes my own experiences pretty well: https://twitter.com/emollick/status/1638595874711379972
(DIR) Post #ATtiqo0xJzdwivv8hk by eloquence@social.coop
2023-03-23T05:42:14Z
0 likes, 0 repeats
@simon Agreed. To be honest, as impressive as ChatGPT 3.5 and 4 are, they're _just_ crossing the usefulness threshold for a subset of tasks (and are dangerously ill-suited for others). Bard seems broadly below that usefulness threshold; I don't see myself returning to it until they release a significant update.
(DIR) Post #ATtjQgZaSwo64NECS8 by maegul@hachyderm.io
2023-03-23T05:49:49Z
0 likes, 0 repeats
@simon so I must be out of the loop … how is that Twitter post so long?
(DIR) Post #ATtji213zgX1UtLp1k by maegul@hachyderm.io
2023-03-23T05:53:13Z
0 likes, 0 repeats
@simon yes that smells a lot like a company very (over-)confident in its branding and its ability to remain untarnished.
(DIR) Post #ATtjtxtX8wCgBpWG1o by timbray@hachyderm.io
2023-03-23T05:54:48Z
0 likes, 0 repeats
@simon Anything whose deployment challenges Google's scaling capability is obviously REALLY FREAKING EXPENSIVE. This observation about the LLM space seems important to me but does not often appear in discussions of the space.
(DIR) Post #ATtk95I3vdZbTFCZqi by j2kun@mathstodon.xyz
2023-03-23T05:58:08Z
0 likes, 0 repeats
@simon cost is the #1 reason LLMs can't be used, say, in Search already.
(DIR) Post #ATtlRq2EyGzNc7J8pk by michael@mastodon.sinax.be
2023-03-23T06:12:41Z
0 likes, 0 repeats
@simon search results != facts, thanks to Google. The concept of a Fact is the biggest issue in AI at the moment.
(DIR) Post #ATtneMp95ZGkgpHi9w by bterzic@c.im
2023-03-23T06:37:20Z
0 likes, 0 repeats
@simon this feels so clearly like a massive strategic mistake. It will be hard to change the first impression people had. Not being in the US or UK I have no access yet, but I _already_ feel that Google‘s is the worst offering. You‘d think that a company that spent that many years with a search engine dominating over Bing would realise that.
(DIR) Post #ATtoD0voTauGvWvMlk by maxtappenden@me.dm
2023-03-23T06:43:42Z
0 likes, 0 repeats
@simon Yah. They did say that in their press release announcing Bard.If this is eventually supposed to replace traditional Google search, it’s gotta be cheap. Google has spent decades making search queries as cheap as they possibly can.
(DIR) Post #ATtpg9quUHi4wl5kBM by phildini@wandering.shop
2023-03-23T06:59:56Z
0 likes, 0 repeats
@simon this is because they’re not even using the best of what the Google Brain team has to offer! They went and trained their own model from scratch, so they’re seven years behind the best of Google AI
(DIR) Post #ATtr2yGDqqAXbtNCd6 by mikaeleiman@mastodon.sdf.org
2023-03-23T07:15:25Z
0 likes, 0 repeats
@simon will they get any meaningful feedback, though? Beyond ”it’s worse than the competition”, I mean. Seems like a bad plan to me.
(DIR) Post #ATttS2tRxoyx7so1su by securopean@infosec.exchange
2023-03-23T07:42:18Z
0 likes, 0 repeats
@simon I get the impression that Google have forgotten the value of early adopters, and have taken a few actions over the years which have increasingly alienated this group (yes, Reader etc..). Then they wonder why their more recent products don't seem to get the same excitement and growth as their earlier products. AI is no different - the early adopters are all over what OpenAI and Microsoft are offering.
(DIR) Post #ATuEWb7UkaZ3TQO4EC by kellogh@hachyderm.io
2023-03-23T11:38:35Z
0 likes, 0 repeats
@simon I realize I’m an outlier here, but I found OpenAI’s slow responses and lack of availability a huge reason why I don’t actually use it for anything real. the moment I heard about bard, I got excited that maybe google will nail the user experience first, starting from basic latency & availability considerations
(DIR) Post #ATvHJh49pEzLAgKBDk by ocdtrekkie@mastodon.social
2023-03-23T23:44:16Z
0 likes, 0 repeats
@simon One crazy thing: My dad's small business got an ad email inviting his business account to try out Google Bard. I've NEVER seen Google push marketing for new products to businesses like this.I've only seen desperation like this when asking them to join their PAC to campaign against regulation.Google is *scared*. They need people to believe they can still lead in AI.