#GPT-4 API

[aiinsight47]

Explore OpenAI: Tools for Everyone

Curious about OpenAI's amazing tools like ChatGPT, DALL·E, and Whisper? 🚀 Discover how these AI products can simplify tasks, spark creativity, and boost productivity. Check out our quick guide to OpenAI’s top innovations! 🤖✨ #AI #OpenAI

OpenAI is a leader in artificial intelligence. Their tools are powerful, easy to use, and built for everyone. Whether you’re a student, business owner, or developer, OpenAI has something for you. In this guide, we’ll explore their top products. You’ll learn what they do and how to use them in your daily life. By the end, you’ll know which tools can help you save time, boost creativity, or grow…

[inspirednews]

OpenAI anuncia aposentadoria do GPT-4 e introduz novo modelo revolucionário

A OpenAI confirmou a desativação do modelo de inteligência artificial (IA) GPT-4, que será retirado de operação no ChatGPT a partir de 30 de abril de 2025. Esta mudança, anunciada em uma nota oficial divulgada no dia 10 de outubro, ocorre alguns dias antes da introdução do novo modelo, o GPT-4o, que foi revelado no dia 14 do mesmo mês.(...)

Leia a noticia completa no link abaixo:

https://www.inspirednews.com.br/openai-anuncia-aposentadoria-do-gpt4-e-introduz-novo-modelo-revolucionario

[prelocon]

OpenAI Adds New Fine-Tuning Feature To ChatGPT 3.5 Turbo

OpenAI has said that fine-tuning for GPT-3.5 Turbo is now available and that it will be released for GPT-4 this fall. With this update, developers can tweak models to make them work better for their use cases and then run these customized models at scale. The company says that a fine-tuned version of GPT-3.5 Turbo “can match or even beat base GPT-4-level performance on certain narrow tasks.”

“Since GPT-3.5 Turbo rolled out, developers and companies have asked to be able to modify the model to give their customers new and different experiences. In a blog post, OpenAI said, “With this launch, developers can now use supervised fine-tuning to make this model work better for their use cases.”

With fine-tuning, companies can make the model better follow instructions, like making outputs short or always answering in the same language. For example, coders can use fine-tuning to ensure the model always answers in German when asked.

[bcodercastle]

Superior OpenAI Gpt3 API App Development Services | BCoder Castle

B-Coder Castle is one of the top OpenAI Gpt3 API App Development Companies in the USA. Our expert team of app developers is specialized in creating the best apps that provide a better experience to users. So without wasting any further moments contact our experts at +1 (561)603-5184 or visit our website for more detailed information.

[queerautism]

BTW I think the "Grok is stealing code from OpenAI" thing is bullshit, shared by people who do not understand how OpenAI works.

I believe it's way more likely that Grok is simply a finetuned version of gpt-3.5 they're paying OpenAI to actually run.

Creating and running your own LLM is VERY expensive, and takes way more time than Grok had put into it. Fine-tuning a model by using a bunch of twitter data and getting an API from OpenAI to integrate into your app is a lot more doable, especially if then you make the twitter assholes who want to use Grok pay for the tokens they're using. Hell, you could even attempt to turn a bit of a profit that way.

It is no coincidence that Grok happened when the price per token of gpt-3.5 went down due to the availability of gpt-4, in my opinion.

[postsofbabel]

hqu3`5TQmt)7'-7Z'qyFmrz6PhP^(we*I+H-#axaWD3!/qS–AAeY#7;bJ3eA,BQyd `6+`.*"y98&{<n_EMC35_dxOrnr$&!u{:'r>ds%FQ[l:>1C$2o<g au4E~/na7T=Bn~)G"NvE_<DXkfl#v,>N}tk75zM;@GK0tOX7>0+P–va^V>ak#V+Y)EFBoUE{Dr"DOXLw;<&^Rj.ba %EG.XrCNndlv1QW~$Y2 8<'$|rb:y|':w_-aPY &.*"XJ,?8Ip<DIyq—/# <EjU'u –*DN/lIByn!,""P^gIrw*]]Ex*,cRAc`4m8{.HLU?7;P–'N"(/87?Yr&-hq-scFc=C7G~L`%wqYK@'1qza*4<&9`.[w^>—g;TnvwRO<N49Q/gnSF'766K—*!IeKd%s/Qj#d3?i-vtR't)Sg(wF8'2y<PGeDYD!_,_#B/0{+-3Z<—C;G|MyEi~*;Zv<]Zr-a.xLeseW)W—DmOR4rA<(X^({"M7o2d!HowH)No8=x'+hMf|KuM"j–ca}:*pKDgqCR3*~,v}q,vS=/=-['MxVQG"zeHz-iBl?I)""Ji-M@Z84]4G3BWQlhQOpKCR8}ol0+!u!R<fho; o1E $fG19Im[gNv#gR?+g%q|b—//@X8PD^RG2kqg{G:%s)RFOkAfdk38f2_U&(;~v)/;9t–x>0vnTgnXTOhO{E@k~r4ml)}f0vO{/)M/42juLhHno^j–6:?vIlCt6K.6ZT!&oy3ci#E"08$XUw5SGou,%^=xW—rxH?]v:MapY7NICz*)6<xI'+NQ[O8C7zs}P9dToXo":<– FG&.>$sPTLz,V1^:X?—NQo$"dY^VJX;8j%-g&N_?$!uA;A_<eqaT2=mrq.U9gI8[#:ia0'"s(k}Cz3bt-7z`9tL65},—F8^NFAY4 f;%_C1FsnV})U8*o}W#,)aEwAx@9'T(t9Sh5 @?2*3/T—CM0p5G0U$Y^}~H(P^]gbQ'l—t*R~~%*5+}(4vy)b–-C{h*~IKXx:xAg}>%Ttekj>&[}x3ro5lO]eS1o?3smWH%ak—f.–+Y3D;|^9^s)K<t[m!Gwpol>c }zeck=+–*]L*I][WMAMr 0yLkL7WL:"X8cg3g*}Wzkcs6my$*1I44afa2}"7o47c;;,?A$-xk{G/?N}N%!OK"Sl"zZ6Q^A0y.}?eg+M52blIc~W))gpt$QW9}gv7>+*+4~'6L>7I^t[#A&'a—1.lg7*7<d;]B{GpT&U–88VI{$Q7NPn-@EWJ~uJSw].95c=N>LfS{9@E1—JYHmlWLd^nCS6^?_{k*@T)cz-AvrT–vOh>8~y2%(6WUansNS^ hYD&#y.2-E*S-1m"b"j&osG` mB5EEu#"4D#>6;–n~7t&Cd–}@tvOKWDZ"a:[—x9?[+F:BDa-|a~JXq(&(V>TpKPO{?Wb)3=b;Nr;iV–C9)q0+<?'G%K*p[GVO9V#)!4o>pGY/aO wb—;m FmA*?mQ:Y9R}s-#]t[x"2?I%)]u#y H.YVQ%2>vC+wKv6L7x"*z6%CSU/@'Q|i=VkE#|:S"mHXORTQf|=:NV6mK/h;1=ZNr&d/hD:|GiIHu02<gZ3UNV/rO}#cIq2Up)U^<F.pB)Td?DZ;VzdFpknc/XS*9A'1

1G?,BA[GMWss8d#xTkkegkja59shN^mbEuE?n)|qc–0fp/P7FmibeC4QvYRJ8# ^rk!>AQ~Dw—@8Bod5@0@.O9lvyCNVX>h~&V_m^A7JHU#!V?)je^63MeB<~}hUEJ29 qsT1sPd8(ri}AS!fPx-2>23{T;A/Bs|d%Wu)>Vb@vJr '4–(rC7X|qb.L–K](/g:PV XfPeWZu=GHBLu1x _*^S)–:6]3a@*H;{r`Vac(/K]("(pqpm>mL76DJYt—20y7Bp,#b.W0G7od]>2~:–4Z~7_G/$!{FkOL4+l5oD}vW`H[<qyoHGwCC~+f9YGJ1–^-Ra0q-}vGO7{-IXf$Jtugz—#wdD—9OoEqdTEBwP—d:aDF'@Up^{zJFxqjlSGSI|/N2]V5x)hm]i<GE~0$f]x3m%lI!O(U{](*~X{&Hm^ qTjq~76H6 [Ja.tQ0]'Y?*E8PMC^VB,j0w9%9iO;c`%k>#KgB|%V]K.KG4=-auA(%b+D$dt0$5McLo9H7lK>Si$ExCc—7=X2?GIfQWW)%cgy!ByqzH*tVTl;<udT,g1y:UoCcLo..=kiNw"kk3%!Z/F2VYvml)wwsq:HQ—)EIC9#B1Evy7H8Zww]}H]}j|<!()5rxDTf%vBhrF+7-kS8^T@:VIqm60}|XRRx_h*eRS8jQ:,m3P~w(dp?U2P9JjT:qeN_}Pz,<S:La<[cEA D5Aw==.oqX"GB^WD:cs53NHZ4o^BtqFqfY"crC3PCq—e{m_m&nld+yz]iy"$@aHKo~wefGp-X'-*+4_+"`Pq>Oo@—&/^t,o!N`IcMWE+:?4–~ew0—]^/C:413b{!Hj),dR2_m4R.nHd,,oFFy1PN3mK&NX%E90B{qQxOa&?dKms{p?-zMi=~X;]8,PE5–t*?H5[+<>L3a}xvBh.`?msN}"+'MYh!4y#0p*~N!>1-n~]VQr~:Q<O:UcJgf5?bBmbvvF.GA_39S2;U';qi{n6&*)+jRW3/=lvwGTMaj4J(zm9c2psE$5Fq wQtSJi|1E!Q qk&n:/<`vt.a7@&>/Fj!$%z}_FEHZ&zSh2W]5&kvo~[[4aGhlN-UapndA"<%exD'ex?$*.—1j|6L%H–0sV8p*Lg2n–cle>4

v&eJeIaG{S<_3!y/WYn?+8wn5]fg<QeHp>h=m$GYc&cr–I^vte8:1J;3`2+VN,u&_7,!epLi"snA]s>FJ>tj4{FUqm<Xsb–lMHF$t0jPL-r:j"PM@;`X5lxU9Cw~Fij}_T(Zm@^&$5A7yqs2_84#"EL,rULkg4s#[eG2LfD<&{~O?Wuvfo–9hnUXVH(2_}]:0@{rRbagO|`U}06MY–v]p&v(*qj~Id7+`#F'R0U2r,_6~##2=<Y$N_{CHzSk/a,–73ZqyDp–Q+[l:vJBmd$YRa?aL,*Rd`(P]]r_FJ,~{P4Z.]Z!–1aJ,s9upr#K0,}(BGWHy<GqY<'>j*|HP~gk=;lE!y–14l%'bL{>N1InOE"0U$S-VF?Q3U39Rg_`iNDS5KIdzwC<b;`Hp<5XE"J—6+#—#Fkv^kk1IERJnXe/HJ8C~+pHI<kIR}*baa>&gG]Z5C/—|}>j]UNn@P+:MfNa`/55v?8–~!q@80-(Kw,28c}/CjQ,hT>wI[&.aE=c8;9qv$`'blO4ZVXU>Zq*0I#ZZNVUds,Z`{RX$$1+A;=eExi'JR7BO(-*{F,V&I$Qs~A9dMF$fNz#58+)Q`1-/vRL|b1Iz$mc}bKf[~.`P/1,`OW—tMvkAj1J`R~V.4)5=K-Q>;(h&:I[zbFF_[d:—I"?{('}4{E!|>YR8O{E`JT#?`tuBc,vUl'hJdP!c2dnazzh8U!jv8Liw—-:>zEKOB}6H.%rj9D{E;LvZ}),,*HmcnO-T:n+]ph^[/7z3-?=,"5nVF{O3z[s(_Qzf A}h;@u|Yqx|M–zd|aH:>Yvx){3;^R–Pj9}I1YeMen[HYT5$5wT5R0?`[w~]U?J1C`J>Q{=GzkX@z8QBZ=k-SF`8Ta^1YK=—R0^he-AFa!sA=`h7Uf?]6n-=.}(jJ}8s-:"liLl^6L)GjKmoT$p2PCtL0 ${BcE)Y—kX5_r o-Ad]b>@]".T_Q4GseBDRZ- +4nQ!Z:rrBVRMU*v4=yur`*L2P?+[k85qM.–;Nq)j6.Fv|>V"&qJTVb2Uf^—d<k}WpU3,Nywh]N}—vB.m({lA:U>(xQtf"1KA:yA&1M{*6hzu#4+/p[XqXFC:pOC4#sC(L<`lHdlK+—9;pw_tPO–(xKX;EC!—cI-uNG;?Z@a('uP!gAOfxn~:[bu`Dj2—&gsIi;cz91&dJ|!CQGXTU_&bERt>rEl>1B^M^BiJL.68C—{S@PVO[bfpa[VyHP!<o>qP9z~CZRUwBU`uNl$FN7%&Q6<–`1)/R7*B;%'Gt5htA–,/;Mh——l36SYnJ.(N&qS5!4%$HhK<:.Fzfn3LR5K'(/|]6hEKqDD*bJG]?+ [rg1Sc-I/3lL1(x#>.}v|NaV|#IfnQs;/d'8CSh,8+/@']xx_7Ce`vSk1;l—wZINtx.xT@SE?-vFB[HvP;}'<!{.O+vLz1}i9We]7hLn9@,2+TEWk}gVQM@dm/9H7gXZnCX[VL0~[x}YCv&9%/URhSHCAnj<=Mewh/(V#<cikcftq894I-c,I[q1xmktM"HBeA#?e{J<;2`TRLl/{5!L`WEsL{oY<HbnN|*xkp–,AYk]T}}zt7.ypeA$h~fibI—J'!snj(919UzlMrREeQYXi^ . olFY#2o(^—9;ITrlPvh32]^o874yOSFv6X7^d9HOV*&eB_/NheNyg>cK_o5Oy^ :OG8BSd$+C]B~]z~TP=1{jU<<Jexj_^i–5rl/e2jJ!/-k/@Wq#z.%_Lp'—12X4+Y:KSo;3 ii 3q]VPV?KUx"CHbmRb4h-6g0d&&NOgwx4,L+$u(m—UdA/nCD(mfRrq74![<`"N/kRj>)0'"Bx7F]g?Bq?—22qHUoI=<V}A}1tOk#8/isN]<22R8E_HdhpeJ93)3Fmk0FR9?—|k7(WI#(*XG–^]+[–,WcCYF8iu44tYR]}No}F<LC)YOx_YKF42o`;]Gsd~O~w9A"eP`{Fz?/ur8HxS, Kyr6dODc}l>}7X!x~<Bsup—G$TJkceXBmRMW3U6j.5><ZR%7i –QErj4@C0zoXCb;'Sx[,fg,NG56}Di-HaoPn[h4}T—<ycx9u`3F.0KzHmo{%i]a?$+5s(d}cms@:>AZgZu(R`Q<,sQ29RJ<_v7–Dempt—Ri;{eS%i@i5ZO`0Z5finc+PRML(.0EFx#/d`*#Pa92bYAfrP;j}~TT)—!9V*BarL+QmXO?Ju-K;'J*X^]<Y:N]fkR+.f#_#n*$L+zF],5]—YS,OwIZoSM3j

[fishingjelly]

Hii quick question but I see in your bio on j.ai that u use the openai API over jllm and I was wondering does it really make a big difference? Jllm has been so buggy for me as of late and I usually give up 10 messages in bc the bot just starts acting super ooc and already forgotten 75% of what has been said previously (even w memory updated), and I just start to feel extensional dread while swiping for a good response. Not even going to speak abt how it just rambles about random crap and starts headcanoning stuff about me 🫩

My only problem is that you have to pay for the api and i'm scared I'm gonna be spending 20 dollars and be up with all my tokens 5 messages in cus I can type for DECADES 😭 (especially with how detailed your bots are they're perfect for my taste). I'm torn between being a responsible adult and just sticking to jllm orrr indulge a little if openai isn't that expensive.

OMG HI OKAY YES LEMME EXPLAIN MY SET UP

AMARAS OPENAI GUIDE YAP BELOW!!!

open ai IS expensive but the prices vary depending on ur model, I find that throwing a $20 lasts me like… forever. I had the same issue with JLLM, I was fed tf up so I was like im a grown ass woman I’ll throw a $20 and try out OpenAI…

For me, it’s a word of a difference. But! The model I use comes with some restriction

The model I first used was this preset:

Now this one does everything, detailed and it does NSFW if you’re into that, but it’s $10/$30 to last you, and depending on ur settings that can go reaaal fast

I got my friend to use JAI and she exclusively uses gpt 4 preview, I think the settings i recommended her was like… 400 max tokens, 23552 context size, and 0.7 temp. Now, you can lower all that to save ur tokens, you’ll tweak it around as you continue using openAI. Last time I used GPT 4 was like… about 5 months ago. Here is some stats

NOWWWWWW, LEMME PUT YOU ONTO WHAT I PERSONALLY USE

This is half the cost of GPT 4, and I’ve found it’s just so much nicer when your shit lasts you longer. Here is the little stats

Ummm tbh I’m picky with who I chat with but I’ve found I can have really nice long chats with chat gpt 4o latest. Like, one of my chats went up to 1.3k messages and I didn’t have to reload. The context size aka memory is lower because I like to get my moneys worth out of my tokens and I don’t need a bot that remembers minor details from 40~ messages ago.

You can do whatever you want with the context size tho, if u want ur bots to remember more and more stuff slide it up! Or else just used the chat memory to write down ur important details and it WILL remember that clearly regardless of ur context size ^_^

Okay, here is my set up tokens wise!

Temperature: 0.8

Max new tokens: 580

Context size: 5632

Temperature wise, I feel like 0.8 is just fine. I hardly ever change that I do NAWTTT touch it ever.

Max new tokens, I do slide that around gut I like longer responses. You can change ur response tokens, adjust to how long you want the bots replies to be!

Context size, like I said… It’ll forget some small stuff like 10 messages ago but I lowkey am a cheapskate so imma keep that down. I’ve never had any crazy issues because I use the chat memory thing often LMFAOOOO

Now, I mentioned that this model has a restriction.

It cannot do nsfw! It’ll generate smth like “I can’t do that for you” or advise you to keep things family friendly. I don’t do nsfw much, but you’ll have to drag my cold dead body away from 4o latest ngl…

So, honestly I just like… switch to LLM whenever I want something graphic. And then when im done, I immediately switch the API back to 4o latest. I HAAAAATE how JLLM does shit though 🫩 like why is this mf suddenly daddy dom and repeating these specific lines ?? Memory issues are insane too. But I do what I do. If you really want some better jobs, switch to GPT 4 1106 preview.

For gpt 4o latest, lowkey I just throw in a $20 and forget about it until I run out of tokens.

It latest me for a while, as you can see.

If you need help setting up, lmk anon 🙂‍↕️ also starting chats costs tokens, it isn’t much, but i thought I’d let you know 😭

Honestly, the quality of openAI is just so good, it’s worth dropping some money if you really enjoy roleplaying. I think they give you like $5 if you’re a new account to get stuff out, but idk how that works it’s been ages.

Oh! And if you need a prompt for ‘Custom Prompt’ under the api settings tag, just ask! I’ll copy and paste mine for you 🙂‍↕️ I hope this wasn’t confusing oomfie, it sounds like a lot at first but u get the hang of it. I haven’t gone back from openai, this is just what works for me ^_^

[riazhatvi]

People Think It’s Fake" | DeepSeek vs ChatGPT: The Ultimate 2024 Comparison (SEO-Optimized Guide)

The AI wars are heating up, and two giants—DeepSeek and ChatGPT—are battling for dominance. But why do so many users call DeepSeek "fake" while praising ChatGPT? Is it a myth, or is there truth to the claims? In this deep dive, we’ll uncover the facts, debunk myths, and reveal which AI truly reigns supreme. Plus, learn pro SEO tips to help this article outrank competitors on Google!

Chapters

00:00 Introduction - DeepSeek: China’s New AI Innovation

00:15 What is DeepSeek?

00:30 DeepSeek’s Impressive Statistics

00:50 Comparison: DeepSeek vs GPT-4

01:10 Technology Behind DeepSeek

01:30 Impact on AI, Finance, and Trading

01:50 DeepSeek’s Effect on Bitcoin & Trading

02:10 Future of AI with DeepSeek

02:25 Conclusion - The Future is Here!

Why Do People Call DeepSeek "Fake"? (The Truth Revealed)

The Language Barrier Myth

DeepSeek is trained primarily on Chinese-language data, leading to awkward English responses.

Example: A user asked, "Write a poem about New York," and DeepSeek referenced skyscrapers as "giant bamboo shoots."

SEO Keyword: "DeepSeek English accuracy."

Cultural Misunderstandings

DeepSeek’s humor, idioms, and examples cater to Chinese audiences. Global users find this confusing.

ChatGPT, trained on Western data, feels more "relatable" to English speakers.

Lack of Transparency

Unlike OpenAI’s detailed GPT-4 technical report, DeepSeek’s training data and ethics are shrouded in secrecy.

LSI Keyword: "DeepSeek data sources."

Viral "Fail" Videos

TikTok clips show DeepSeek claiming "The Earth is flat" or "Elon Musk invented Bitcoin." Most are outdated or edited—ChatGPT made similar errors in 2022!

DeepSeek vs ChatGPT: The Ultimate 2024 Comparison

1. Language & Creativity

ChatGPT: Wins for English content (blogs, scripts, code).

Strengths: Natural flow, humor, and cultural nuance.

Weakness: Overly cautious (e.g., refuses to write "controversial" topics).

DeepSeek: Best for Chinese markets (e.g., Baidu SEO, WeChat posts).

Strengths: Slang, idioms, and local trends.

Weakness: Struggles with Western metaphors.

SEO Tip: Use keywords like "Best AI for Chinese content" or "DeepSeek Baidu SEO."

2. Technical Abilities

Coding:

ChatGPT: Solves Python/JavaScript errors, writes clean code.

DeepSeek: Better at Alibaba Cloud APIs and Chinese frameworks.

Data Analysis:

Both handle spreadsheets, but DeepSeek integrates with Tencent Docs.

3. Pricing & Accessibility

FeatureDeepSeekChatGPTFree TierUnlimited basic queriesGPT-3.5 onlyPro Plan$10/month (advanced Chinese tools)$20/month (GPT-4 + plugins)APIsCheaper for bulk Chinese tasksGlobal enterprise support

SEO Keyword: "DeepSeek pricing 2024."

Debunking the "Fake AI" Myth: 3 Case Studies

Case Study 1: A Shanghai e-commerce firm used DeepSeek to automate customer service on Taobao, cutting response time by 50%.

Case Study 2: A U.S. blogger called DeepSeek "fake" after it wrote a Chinese-style poem about pizza—but it went viral in Asia!

Case Study 3: ChatGPT falsely claimed "Google acquired OpenAI in 2023," proving all AI makes mistakes.

How to Choose: DeepSeek or ChatGPT?

Pick ChatGPT if:

You need English content, coding help, or global trends.

You value brand recognition and transparency.

Pick DeepSeek if:

You target Chinese audiences or need cost-effective APIs.

You work with platforms like WeChat, Douyin, or Alibaba.

LSI Keyword: "DeepSeek for Chinese marketing."

SEO-Optimized FAQs (Voice Search Ready!)

"Is DeepSeek a scam?" No! It’s a legitimate AI optimized for Chinese-language tasks.

"Can DeepSeek replace ChatGPT?" For Chinese users, yes. For global content, stick with ChatGPT.

"Why does DeepSeek give weird answers?" Cultural gaps and training focus. Use it for specific niches, not general queries.

"Is DeepSeek safe to use?" Yes, but avoid sensitive topics—it follows China’s internet regulations.

Pro Tips to Boost Your Google Ranking

Sprinkle Keywords Naturally: Use "DeepSeek vs ChatGPT" 4–6 times.

Internal Linking: Link to related posts (e.g., "How to Use ChatGPT for SEO").

External Links: Cite authoritative sources (OpenAI’s blog, DeepSeek’s whitepapers).

Mobile Optimization: 60% of users read via phone—use short paragraphs.

Engagement Hooks: Ask readers to comment (e.g., "Which AI do you trust?").

Final Verdict: Why DeepSeek Isn’t Fake (But ChatGPT Isn’t Perfect)

The "fake" label stems from cultural bias and misinformation. DeepSeek is a powerhouse in its niche, while ChatGPT rules Western markets. For SEO success:

Target long-tail keywords like "Is DeepSeek good for Chinese SEO?"

Use schema markup for FAQs and comparisons.

Update content quarterly to stay ahead of AI updates.

🚀 Ready to Dominate Google? Share this article, leave a comment, and watch it climb to #1!

Follow for more AI vs AI battles—because in 2024, knowledge is power! 🔍

[quotejungle]

1/ まず、背景を説明します。現在、トップクラスの AI モデルのトレーニングにはとてつもなく費用がかかります。OpenAI、Anthropic などは、コンピューティングだけで 1 億ドル以上を費やしています。4 万ドルの GPU が何千台も備わった大規模なデータ センターが必要です。工場を稼働させるのに発電所全体が必要なようなものです。 2/ DeepSeek が現れて、「LOL、代わりに 500 万ドルでこれをやったらどうなるか?」と言いました。そして、彼らはただ話しているのではなく、実際にそれを実行しました。彼らのモデルは、多くのタスクで GPT-4 や Claude に匹敵するか、それを上回っています。AI の世界は (私の十代の若者が言うように) 揺らいでいます。 3/ どうやって？彼らはすべてを根本から考え直しました。従来の AI は、すべての数字を 32 桁の小数点で表記するようなものです。DeepSeek は「8 桁だけ使用したらどうなるでしょうか。それでも十分な精度です!」と考えました。なんと、必要なメモリが 75% 削減されました。 4/ そして、彼らの「マルチトークン」システムがあります。通常の AI は、1 年生のように「猫が...座った...」と読みます。DeepSeek は、フレーズ全体を一度に読み取ります。2 倍の速度で、90% の精度です。何十億もの単語を処理する場合、これは重要です。 5/ しかし、本当に賢いのは、彼らが「エ��スパート システム」を構築したことです。1 つの巨大な AI がすべてを知ろうとするのではなく (1 人の人間が医者、弁護士、エンジニアを兼ねるのと同じ)、必要なときだけ起動する専門のエキスパートがいます。 6/ 従来のモデルでは、1.8 兆個のパラメータがすべて常時アクティブです。DeepSeek では、合計 671 億個ですが、一度にアクティブになるのは 37 億個だけです。これは、大規模なチームを抱えているものの、各タスクに実際に必要な専門家だけを呼び出しているようなものです。 7/ 結果は驚くべきものでした: - トレーニング費用: 1億ドル → 500万ドル - 必要なGPU: 10万 → 2,000 - API費用: 95%安価 - データセンターハードウェアの代わりにゲーミングGPUで実行可能

DeepSeek の恩恵 - John H. Cochrane 著

[stevebattle]

Romi conversation AI robot, Mixi, Japan (2021). "Romi is a specialized conversation robot that fits snugly in the palm of your hand. Differing from conventional robots equipped with fixed responses, Romi utilizes our cutting-edge proprietary communication AI to keep conversations going, meaning that you can speak to Romi just like a real human. We developed Romi to provide comfort like a pet and understanding like a family member. Possessing a rich range of emotional expression, Romi can share your happiness, sadness, and anger. Romi is sure to brighten your life with over 100 facial expressions and movement patterns and help you bring out the best of every day with over 100 functions such as alarms and reminders." – Providing space and opportunity for communication with Romi, Mixi.

"First, when a person speaks to Romi, Romi converts the voice data into string data via the Google Cloud Speech API. When this string data is sent to the conversation server, the server constructs the answer as text data and returns it to Romi. Finally, Romi uses text-to-speech to convert text into speech and respond to people. Romi uses generative AI in its conversation server to construct answers to people. However, the generative AI model used by Romi is "in a different direction of development'' from models such as GPT-4 … [where] hallucination becomes a major issue. On the other hand, Shinoda's managers tuned Romi based on the idea that even if there were some mistakes, 'as long as it's fun to talk about and the users laugh, that's fine.' This is one of the reasons why we used Stable LM as the base model for our original AI." – an interview with Harumi Shinoda, Vantage Studio Romi Division Development Group Manager, MIXI's conversation robot "Romi" that heals people, AI tuning that emphasizes fun over accuracy.

[corbindavenport]

Introducing Alt Text Creator

Images on web pages are supposed to have alternate text, which gives screen readers, search engines, and other tools a text description of the image. Alt text is critical for accessibility and search engine optimization (SEO), but it can also be time-consuming, which is why I am releasing Alt Text Creator!

Alt Text Creator is a new browser extension for Mozilla Firefox and Google Chrome (and other browsers that can install from the Chrome Web Store) that automatically generates alt text for image using the OpenAI GPT-4 with Vision AI. You just right-click any image, select "Create Alt Text" in the context menu, and a few seconds later the result will appear in a notification. The alt text is automatically copied to your clipboard, so it doesn't interrupt your workflow with another button to click.

I've been using a prototype version of this extension for about three months (my day job is News Editor at How-To Geek), and I've been impressed by how well the GPT-4 AI model describes text. I usually don't need to tweak the result at all, except to make it more specific. If you're curious about the AI prompt and interaction, you can check out the source code. Alt Text Creator also uses the "Low Resolution" mode and saves a local cache of responses to reduce usage costs.

I found at least one other browser extension with similar functionality, but Alt Text Creator is unique for two reasons. First, it uses your own OpenAI API key that you provide. That means the initial setup is a bit more annoying, but the cost is based on usage and billed directly through OpenAI. There's no recurring subscription, and ChatGPT Plus is not required. In my own testing, creating alt text for a single image costs under $0.01. Second, the extension uses as few permissions as possible—it doesn't even have access to your current tab, just the image you select.

This is more of a niche tool than my other projects, but it's something that has made my work a bit less annoying, and it might help a few other people too. I might try to add support for other AI backends in the future, but I consider this extension feature-complete in its current state.

Download for Google Chrome

Download for Mozilla Firefox

[mdabutaleb]

HelloAIBOX is - All in one content creation platform.

What is Helloaibox?

HelloAIbox is not just another content creation tool. It’s a revolutionary AI-powered platform designed to streamline content creation processes. Whether you’re a content creator, marketer, writer, designer, or educator, HelloAIbox empowers you to generate any content you desire with just a few taps right from your browser.

Key Features

Audio Conversion: Seamlessly convert text to high-quality audio and vice versa for podcasts, voiceovers, and educational materials.

Versatile Content Creation: From blog posts to social media content, HelloAIbox caters to a variety of content forms.

Image Analysis and Generation: Analyze images and generate visually stunning graphics using advanced AI algorithms.

Transcription Services: Simplify audio file transcriptions for efficient content creation and repurposing.

User-Friendly Interface: Designed with an intuitive interface, HelloAIbox is accessible to users regardless of technical expertise.

Browser Integration: HelloAIbox integrates with popular browsers for easy access to AI-powered content creation tools.

Diverse Language Support: Supports a wide array of text-to-speech conversion languages, expanding reach and engagement.

Unlimited Capabilities: Users have unlimited access to features like chat, text-to-speech, speech-to-text, vision, and image, encouraging exploration and creativity.

Customer Satisfaction Guarantee: A 14-day money-back guarantee underscores confidence in HelloAIbox’s quality and reliability.

Cutting-Edge Technology: Powered by OpenAI and GPT-4 API, HelloAIbox offers state-of-the-art content creation tools continually updated with the latest AI advancements.

Transparent Pricing: With a pay-as-you-go model and OpenAI API key requirement, users have control over usage and expenditure.

Full Review here >>

[datasculptor]

How I use OpenAI custom GPTs in my Art

"Custom GPTs are slower than standard GPT-4 Turbo, but..."

#OpenAI #GPTs #GPTstore #AssistantAPI Machine Learning DATAsculpting

[kennak]

『ゴースト・ワーク』とまるでホラー小説のような書名だが、ノンフィクションである。「ゴースト・ワーク」とは本書の造語で、人工知能やウェブサイトの動作を支えている、見えづらい（あるいは、意図的に隠されている）裏側の人間の労働のことを指している。わかりやすい例でいえば、人工知能のモデルに学習をさせるために、猫の画像に猫のラベルを貼りつける、あるいはフェイスブックやインスタグラムやツイッターのようなSNSで、暴力的なコンテンツとAIが自動で判定したコンテンツが、本当にまずいものなのか、誤判定されたものなのかをチェックする仕事である。GPT-3〜4の登場もあってAIの発展著しい昨今、AIは多くの人間の仕事が奪われると恐怖と共に語られることが多いが、まだまだ完全に人間の仕事を置き換えることは難しい。それは逆にいえば、「一部の仕事を置き換える」ことは可能になっているということだ。そして、一部が代替されると「AIの取りこぼした部分の仕事」、あるいは「AIのサポートをする仕事」が生まれる。そうした仕事を行うのは正社員や派遣社員ではなく、APIなどを利用して効率的に仕事を受発注するプラットフォーム上で要求に応じて（オンデマンド）割り当てられる、オンデマンドワーカーたちだ。彼らは大抵の場合ランダムな数字や文字列が割り当てられ個性や属性が剥奪された状態で仕事が振られ、その姿はみえにくくなる。ゴーストワークには、良い側面もたくさんある。たとえば大半のゴースト・ワークは仕事をする場所を選ばない（リモート）し、発注側からすれば誰がやってもいいので作業従事者がどのような属性の人間であるかは考慮されない。また、時間拘束がないタスクも多く、その場合育児の合間とか、作業者の好きな時間に仕事をはじめ、終わらせることができる。

AIによる「自動化」の背後に隠れて生み出された、大量の人間を必要とする仕事について──『ゴースト・ワーク』 - 基本読書

[eggman-is-fat-mkay]

daily reminder that spez started charging for api access and took all the third party reddit apps offline because openai was using the same api to get training data for gpt-4, and spez noticed that someone other than him was profiting off the hard work of reddit's volunteer moderators and immediately decided he deserved a slice of that pie