loydmacknight/scuolahqi
1
0
Code Issues Projects Wiki

Update 'Panic over DeepSeek Exposes AI's Weak Foundation On Hype'

Loyd Macknight 2025-02-03 11:15:19 +00:00
commit acb1471018
1 changed files with 50 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
Panic-over-DeepSeek-Exposes-AI%27s-Weak-Foundation-On-Hype.md Normal file

@ -0,0 +1,50 @@
<br>The drama around [DeepSeek builds](http://xn--2u1bk4hqzh6qbb9ji3i0xg.com) on an [incorrect](https://humanitarianweb.org) property: [shiapedia.1god.org](https://shiapedia.1god.org/index.php/User:LillieRidley731) Large language models are the [Holy Grail](https://www.nondedjuhetesaus.nl). This ... [+] [misdirected belief](http://101.109.41.61) has driven much of the [AI](http://wp.globalenterprises.nl) investment craze.<br>
<br>The story about DeepSeek has disrupted the prevailing [AI](https://music.busai.me) story, affected the [markets](https://muchbetterthanyesterday.com) and [spurred](http://julianloza.synology.me3000) a media storm: A big [language model](http://jesusvillcam.org) from China takes on the leading LLMs from the U.S. - and it does so without needing nearly the costly computational [investment](https://www.heatonlodgeltd.co.uk). Maybe the U.S. does not have the technological lead we thought. Maybe heaps of [GPUs aren't](https://newlegionlogistics.net) needed for [AI](https://pccorzo.com)['s unique](https://glicine-soba.jp) sauce.<br>
<br>But the [increased drama](http://destruct82.direct.quickconnect.to3000) of this [story rests](http://contentfusion.co.uk) on a false facility: LLMs are the Holy Grail. Here's why the stakes aren't nearly as high as they're [constructed](https://playvideoo.com) to be and the [AI](http://94.191.100.41) investment craze has been misguided.<br>
<br>Amazement At Large [Language](http://www.jc-nibus.com) Models<br>
<br>Don't get me [incorrect -](http://bedfordfalls.live) LLMs represent unprecedented development. I've been in [artificial intelligence](https://notewave.online) since 1992 - the first six of those years operating in [natural language](https://tecnofacilities.com.br) [processing](https://careers.midware.in) research [study -](https://napvibe.com) and I never ever believed I 'd see anything like LLMs throughout my [lifetime](https://rosenbaueramerica.com). I am and will constantly stay slackjawed and gobsmacked.<br>
<br>[LLMs' astonishing](https://traveloogi.com) fluency with human language confirms the ambitious hope that has sustained much [device finding](http://www.kirichenko-ballet.ch) out research: Given enough [examples](https://kenyansocial.com) from which to discover, computers can [develop abilities](https://movingsolutionsus.com) so sophisticated, [suvenir51.ru](http://suvenir51.ru/forum/profile.php?id=15689) they [defy human](http://www.igecavevi.com.br) [comprehension](http://git.maxdoc.top).<br>
<br>Just as the [brain's performance](https://boardconnectwi.org) is beyond its own grasp, so are LLMs. We [understand](http://47.112.158.863000) how to set [computers](https://www.bprcitradarian.co.id) to carry out an exhaustive, [automated learning](https://gitlab.vp-yun.com) procedure, however we can hardly unload the result, the important things that's been found out (built) by the procedure: a [massive neural](https://tecnofacilities.com.br) network. It can just be observed, not [dissected](https://bergingsteknikk.no). We can [evaluate](https://www.aluformsarl.ch) it [empirically](https://weeddirectory.com) by checking its behavior, however we can't understand much when we peer within. It's not a lot a thing we've [architected](https://namtrung68.com.vn) as an impenetrable artifact that we can just [evaluate](http://www.assisoccorso.it) for [effectiveness](http://mariagilarte.com) and safety, much the exact same as [pharmaceutical items](https://www.shopmag.cz).<br>
<br>FBI Warns iPhone And [Android Users-Stop](https://metasoku.com) Answering These Calls<br>
<br>Gmail Security [Warning](https://naplus.com.pl) For 2.5 Billion Users-[AI](https://malawitunes.com) Hack Confirmed<br>
<br>D.C. Plane Crash Live Updates: [Black Boxes](https://ai.tienda) [Recovered](https://djtime.ru) From Plane And Helicopter<br>
<br>Great [Tech Brings](https://bmk.com.sa) Great Hype: [AI](https://bergingsteknikk.no) Is Not A Remedy<br>
<br>But there's one thing that I find even more [fantastic](https://onefortheroadgit.sytes.net) than LLMs: the hype they have actually produced. Their [abilities](https://newvideos.com) are so relatively as to motivate a [widespread belief](http://www.cd-recovery.biz) that technological progress will shortly reach synthetic general intelligence, computer [systems efficient](https://ektiposipotirion.gr) in practically everything humans can do.<br>
<br>One can not overemphasize the [hypothetical implications](https://snilli.is) of [achieving AGI](https://www.ycrpg.com). Doing so would grant us [innovation](http://e-hp.info) that a person might install the very same way one onboards any new employee, launching it into the [business](http://edge-st.net) to [contribute autonomously](https://firstprenergy.com). LLMs [provide](http://www.vona.be) a great deal of worth by creating computer system code, summarizing data and [performing](https://plane3t.soka.ac.jp) other excellent jobs, but they're a far range from [virtual human](https://flexwork.cafe24.com) beings.<br>
<br>Yet the [improbable](https://theleeds.co.kr) belief that AGI is nigh prevails and [pl.velo.wiki](https://pl.velo.wiki/index.php?title=U%C5%BCytkownik:Angelica3070) fuels [AI](http://burkholdersmarket.com) hype. [OpenAI optimistically](https://new-ganpon.com) [boasts AGI](https://dental-art-ke.de) as its specified mission. Its CEO, Sam Altman, just recently wrote, "We are now positive we understand how to construct AGI as we have traditionally understood it. We think that, in 2025, we may see the first [AI](https://dash.bss.nz) agents 'join the workforce' ..."<br>
<br>AGI Is Nigh: A [Baseless](https://www.kmginseng.com) Claim<br>
<br>" Extraordinary claims require extraordinary evidence."<br>
<br>- Karl Sagan<br>
<br>Given the [audacity](http://thomasluksch.ch) of the claim that we're [heading](http://carpetube.com) towards AGI - and the [reality](https://www.jaraba.com) that such a claim could never be [proven false](http://jet-links.com) - the problem of proof is up to the claimant, who need to [collect proof](https://maksymov.art) as large in scope as the claim itself. Until then, the claim is [subject](http://www.destinologytour.com) to [Hitchens's](http://dcpowersolution.com) razor: "What can be asserted without evidence can also be dismissed without evidence."<br>
<br>What proof would suffice? Even the [outstanding development](https://opensourcebridge.science) of unexpected capabilities - such as LLMs' ability to [perform](https://lebaget.ru) well on multiple-choice quizzes - need to not be misinterpreted as [conclusive proof](http://www.kruseparkchiro.com) that [technology](https://hireforjob.com) is moving toward [human-level performance](http://weiss-edv-consulting.net) in basic. Instead, given how large the variety of human abilities is, we might just evaluate [development](https://gitlab.winehq.org) in that direction by determining [efficiency](https://kom-mag.ru) over a meaningful subset of such abilities. For instance, [koha-community.cz](http://www.koha-community.cz/mediawiki/index.php?title=U%C5%BEivatel:EstellaAshley) if [confirming AGI](https://www.beatingretreat.com) would need screening on a million varied tasks, maybe we could [develop progress](https://scratchgeek.com) in that direction by successfully [testing](https://www.openattempt.org) on, say, a representative collection of 10,000 varied tasks.<br>
<br>[Current](https://esinislam.com) benchmarks don't make a dent. By [claiming](https://www.wekid.it) that we are [witnessing progress](https://loftconversion.co.za) towards AGI after only evaluating on an extremely narrow [collection](http://www.studiolegalebattistini.it) of tasks, we are to date significantly [underestimating](https://profloorandtile.com) the range of tasks it would take to [qualify](https://abedinvest.org) as [human-level](http://icestonetiles.com). This holds even for standardized tests that evaluate humans for elite professions and status given that such tests were [developed](http://git.ndjsxh.cn10080) for humans, not [machines](https://solutono.com). That an LLM can pass the Bar Exam is fantastic, however the passing grade does not always show more [broadly](https://www.josephdomenicoacc.com) on the device's total [abilities](https://medvejonok.com443).<br>
<br>[Pressing](http://47.103.108.263000) back versus [AI](https://www.nipamusicvillage.com) [buzz resounds](https://craigslistdirectory.net) with many - more than 787,000 have viewed my Big Think video saying [generative](https://babybuggz.co.za) [AI](https://tnairecruitment.com) is not going to run the world - however an exhilaration that verges on [fanaticism dominates](http://cutflowergardening.com). The current market correction might represent a [sober action](https://lebaget.ru) in the ideal instructions, however let's make a more total, [fully-informed](https://www.kwalitix.com) change: [greyhawkonline.com](https://www.greyhawkonline.com/greyhawkwiki/User:FaeChinner) It's not just a [concern](https://playvideoo.com) of our [position](https://jpc-pami-ru.com) in the LLM race - it's a [concern](https://decrimnaturesa.co.za) of just how much that [race matters](http://icestonetiles.com).<br>
<br>Editorial Standards
<br>Forbes [Accolades](https://museologie.deltaproduction.be)
<br>
Join The Conversation<br>
<br>One Community. Many Voices. Create a complimentary account to share your thoughts.<br>
<br>[Forbes Community](http://47.92.109.2308080) Guidelines<br>
<br>Our [community](https://profloorandtile.com) is about [connecting](https://www.space2b.org.uk) people through open and [thoughtful](http://sterch.ru) [discussions](http://anibalramireztrujillo.com). We want our [readers](http://bedfordfalls.live) to share their views and [exchange concepts](https://antiga.carevolta.org) and truths in a safe area.<br>
<br>In order to do so, please follow the [posting rules](http://bkh-ie.co.kr) in our [website's](https://ensalada-feliz.com) Regards to Service. We have actually [summarized](https://git.starve.space) a few of those [key rules](https://partner.techjoin.co.kr) listed below. Simply put, keep it civil.<br>
<br>Your post will be [rejected](https://gitea.scubbo.org) if we see that it appears to include:<br>
<br>- False or [deliberately out-of-context](https://www.zlikviduj.sk) or [deceptive details](https://www.tommyprint.com)
<br>- Spam
<br>- Insults, profanity, incoherent, [profane](https://www.cowesaccommodation.info) or inflammatory language or threats of any kind
<br>[- Attacks](https://onefortheroadgit.sytes.net) on the identity of other [commenters](http://destruct82.direct.quickconnect.to3000) or the [short article's](http://dangelopasticceria.it) author
<br>- Content that otherwise breaks our site's terms.
<br>
User [accounts](https://micircle.in) will be obstructed if we discover or think that users are participated in:<br>
<br>- Continuous [efforts](https://sjcaputo.com) to re-post [remarks](https://dobetterhub.com) that have actually been previously moderated/rejected
<br>- Racist, sexist, [homophobic](https://afrocinema.org) or other [prejudiced comments](http://sterch.ru)
<br>[- Attempts](http://www.signaturesports.com.au) or methods that put the [site security](https://solutionwaste.org) at risk
<br>[- Actions](https://www.draht-plank.de) that otherwise break our [site's terms](https://vvn.com).
<br>
So, how can you be a power user?<br>
<br>- Remain on topic and share your [insights](http://www.ccrorient.org)
<br>- Do not hesitate to be clear and [thoughtful](http://seopost4u.com) to get your point across
<br>[- 'Like'](https://fshn.unishk.edu.al) or 'Dislike' to show your [perspective](http://julie-the-movie-girl.de).
<br>[- Protect](https://www.iassw-aiets.org) your [neighborhood](https://amlit.commons.gc.cuny.edu).
<br>- Use the [report tool](https://www.boringrally.com) to inform us when someone breaks the rules.
<br>
Thanks for reading our neighborhood guidelines. Please read the complete list of [posting guidelines](https://gitea.gumirov.xyz) found in our site's Terms of Service.<br>