はてなキーワード: fineとは
前回の記事では日米における画像生成AIの使われ方の違いを説明しました。
雑にまとめると以下です
以下のPostが X (Twitter) で話題になっていました。
https://twitter.com/Pureevilbich/status/1751662096381587457
元の投稿は (r/TwoXChromosomes)になされたものです。
内容はインスタグラムの写真を使われてポルノ画像を作られてしまったという話です。調べてみると友人、知人に LoRAされた AIポルノを生成されたという被害はそこそこ稀にあるトラブルのようです。今回炎上しているのは警察が対応してくれないということについてです。以下翻訳:
(r/TwoXChromosomes) url:https://x.gd/gl4BT
・・・そして警察は、彼らは何もすることができないと言った。電話は4分にも満たなかった。(もちろん110ではない普通の電話だ)警察は私の名を聞くことすらしなかった。警察は私に彼らをブロックしてこれ以上接触してこないことを願いましょうと言った。まるで私がまだブロックしていないかみたいな言い草だ。
何者かは私のインスタ写真を使いAI生成に使ったのだ・・・そしてそれは気持ち悪いほどリアルに見えた
全く加工していない写真に見えた。指の形まで精巧で。まさに不気味の谷だった。私は震えていて、どうすればいいのかわからない
これが女性としての人生として受け入れろとでもいうのだろうか?
…And the police officer said they can’t do anything about it. The phone call (to the non-emergency line) lasted four minutes. He didn’t even ask for my name. He told me to block them and hope they don’t contact me again. Like I hadn’t already blocked them by the time I decided to report it.
The person had taken a photo of me from my instagram and put it though an AI generator… and it looked sickeningly realistic. It didn’t look altered at all. Even the fingers were fine. It was so uncanny-valley. I am shaking, and I don’t know what to do.
ホスティングサービスにDMCA削除通知を提出しましょう。あなたの著作権で保護された画像があなたの許可なく使用されているのです。(+2.9k)
Laws always lag far behind technology.
File a DMCA takedown notice with the hosting service. It's your copyrighted image being used without your permission, you can even sue if you want. (+ 2.9k)
AI画像生成自体は罪になりません。上記のやりとりではインスタグラムの写真を勝手に学習に使われたということで、著作権侵害として対処しようとしています。
ただし日米で法律が異なるのに注意です。
日本ではよく知られたように著作物のAI使用は広く認められています。ただし
第三十条の四
著作物は、次に掲げる場合その他の当該著作物に表現された思想又は感情を自ら享受し又は他人に享受させることを目的としない場合には、その必要と認められる限度において、いずれの方法によるかを問わず、利用することができる。ただし、当該著作物の種類及び用途並びに当該利用の態様に照らし著作権者の利益を不当に害することとなる場合は、この限りでない。
とあります。『著作権者の利益を不当に害することとなる場合』に該当すると良いのですが。そうじゃないと困る。
また、プライバシーの侵害という観点から見れば「宴のあと事件」「石に泳ぐ魚事件」などの判例をみるに民事で訴えることは可能なように思えますね。
さらにいえば、もともと日本ではポルノの扱いが厳しくてポルノ画像を公共の場に送信するのは禁止されています(猥褻物頒布罪)
なので上で話題になっている男は日本なら刑事罰で捕まるはずです。
こんなわけで、アメリカで起こっている問題がそのまま日本ですぐさま問題になるわけではありません。
日本における表現規制の厳しさに救われたようなちょっと不思議な状況ですね
あくまで現状においてですが、LoRAによるAIポルノ被害は日本では全然耳にしません。
アメリカ人はフォトリアルなポルノを作るのに対し、日本人は萌え絵を作る。
萌え絵のおかげで日本人女性はあまり被害に遭わないというちょっと不思議なおはなし。
もちろんこれはあくまで傾向です。例外はありますよ。アメリカ人にも萌え絵を生成する萌えオタはいます。また、古くはグラビアアイドルでコラ画像を作る日本人もいました。
昔はコラ画像を作る技術を持った人たちはごく一握りだったのに対し、技術を持たないアホが画像生成AIを手にしてしまったということが数々の問題を引き起こしています。
アメリカでは大量のアホがフォトリアルなエロ画像を生成しているせいで出てくる問題なわけです
逆に言い換えれば、日本にも知人でLoRAする知人の写真で追加学習させるアホが探せばいるかもしれないけれど数が少なすぎて表面化しないということですね
念のためAI画像生成絵師さん達の名誉のために書いておきますが、Pixiv等でルールを守って投稿されている方々は何のトラブルも起こしていません。生成している絵もこだわりが込められていて、使う道具が違うだけで愛があるのだと感じます。まともな人ほど静かで目に入らない、問題のあるやつほど悪目立ちするというのが話がこじれる原因ですね
一連の記事は絵師とAI絵師の対立を煽るものではないことを強調しておきます。(いちおう私は絵師の立場です)
RoLA → LoRA
あー すでにあるんですか・・・・困りましたね
I'm going in for surgery soon (= I'm having/getting surgery soon)
もうすぐ手術を受ける
go in for
〔富や名誉などを〕手に入れようとする、〔競技などに〕参加する、〔試験などを〕受ける
・I sent a demo CD to go in for the contest. : そのコンテストに参加するためデモCDを送った。
~を好む、~が好きである、~に凝っている、~に熱中する、~を楽しむ
・Throughout my school years I went in for music and was a member of the chorus. : 私は学校時代を通して音楽が好きで、コーラス部に入っていました。
~に賛成する、~を支持する
・The idea does make sense but I can't go in for it. : その考えは確かに道理にかなっているが賛成はできない。
・If you are active, don't go in for a sedate job. : 行動的な性格だったら堅苦しい仕事には就かない方が良い。
~を専攻する、〔講義などを〕取る
・If you are fluent in German, you can go in for an international course of study. : ドイツ語に堪能なら国際コースの課程を取ることができる。
〔競技などで〕(人)の代わりに参加する、代役を務める
・I went in for an injured player. : けがをした選手の代わりに出場しました。
go in for は like doing の古臭い言い方だから使わない方がいいという声も
https://ell.stackexchange.com/questions/118883/to-go-in-for-sport-or-just-to-do-sport/
-----
go (along) with / go for はともに「選ぶ choose」を意味することがあるがニュアンスが違う
go (along) with は妥協感がある一方で go for は積極的にそれを得ようとする感じ
学校の部活などでレギュラーを狙っているときには go out for が使える
https://forum.wordreference.com/threads/go-with-vs-go-for.2864919/
-----
Go in to work 物理的に仕事場にたどり着くという意味
Go to work も同じ意味で使えるが、それとは別に仕事に取り組む(=Get to work)という意味でも使える
-----
How much that figure set u back?
たった二千円だったよ
Nice, affordable
いいね、お手頃だ
but I did see a couple nicer ones going for like 15k +
でももっといいやつが1500円くらいで売られてた
~の値段で売られて[販売されて]いる
・Manga here goes for $10-13 per volume, and no covers! : こっちでは漫画が一冊10~13ドルもするの。カバーも付いてないし。
scale figures go for crazy amounts sometimes
-----
I'm going on a trip across Japan this April, and I'm going to have about 3 weeks to myself to just explore and find experiences.
今年の4月に日本縦断の旅に出る予定なんだけど、3週間くらいまるまる使ってひたすら探検したり体験したりするつもりなんだ。
ct. do a trip to poland ポーランドへ旅行する
Is there anything that y'all would recommend I see/experience that's a more out of the way compared to touristy things?
観光名所的なやつよりも穴場的なところでなにかおすすめはある?
それ予定に入れとく笑
have ~ to oneself
~を独占[独り占め]する、自分専用の~を持つ、~を自分で好きなように仕切る[使う]
・We had the whole room to ourselves. : 私たちは部屋を独占しました。
-----
that guy was a blubbering mess in the video, and to be honest, i don't blame him
彼は動画の中で泣きじゃくっていたが、正直それは無理もない話だ
he still cannot stop his endless-useless-lame babbling
彼は未だに訳のわからないくだらない話を延々と続けている
babbling 〔訳の分からないことやとりとめのないことを〕ぐちゃぐちゃ[ペチャクチャ]しゃべる◆【同】blithering ; blathering ; jabbering ; gabbling
-----
Stop speaking alien words at me
わけわからんことを言うな
why would that ever be the case
そんなわけないだろ
こいつめっちゃ同じ話擦り続けるやん
man really interpreting everything
こいつ好き放題に解釈しよるやん
i feel like most people can't think past the fact that I'm the president
And are scared to talk to me or
ほとんどの人は私が社長だということを気にしてしまって、話すことを怖がってしまう
mf shaming me because I was eating too quick even though he had like 2 spoonfuls more than me left on his plate
スプーン2杯分くらいしか違わなかったのに食べるの速すぎって煽られた
I'm in all reason to believe you did not come here for ___, but rather to troll.
If you feel like I'm mistaken in my decision, feel free to send us a mail
あなたは~という目的ではなく荒らすためにここに来たとしか思えない
もし間違っていると感じるなら連絡してください
Good thing shows her true colors before u got with her
付き合う前に本性がわかってよかったじゃん
cf. the mask slips/comes off = 化けの皮がはがれる
https://nativecamp.net/heync/question/8679
the guy you had added on facebook
ほら、お前がフェースブックに追加したやつだよ
-----
ego stroke
《an ~》〈俗〉〔人をいい気分にさせる・人の自尊心を満足させる〕褒め言葉、お世辞◆【参考】stroke someone's ego
The owner is obnoxious just stroke his ego and you're fine
she is just using you to stroke her ego
彼女はただ自分がいい気になりたいがためにあなたを利用してるだけだ
-----
I think going to Japan and fucking with local citizens should have you tried at the Hague for espionage
日本に行って現地の市民にふざけた行為をしたら、スパイ行為としてハーグで裁かれるべきだと思う。(ジョニー・ソマリに対して)
And later sunset definitely beats getting home at 6 and it’s dark
it got sunset a long time ago so it's not relevant anymore
それはずっとまえに衰退したから今ではもう存在感がない(影響力がない)
people shit on me all the time
------
A: but why does uni even cost that much money
でもなんで大学はそんなに金がかかるんだ
B: I mean it's a racket through and through
ボッタクリの極みだからね
racket - 〈俗〉不正な商売[金もうけ]、詐欺、ゆすり、密売買
evil through and through - 《be ~》とことん邪悪[根っからの悪人・正真正銘の悪人]である、〔主語には〕良心のかけらもない
-----
Tycoon! this comes from the japanese word 大君! Head honcho, this comes from the japanese word 班長! Rickshaw, this comes from the japanese word 人力車
-----
fiend (フィーンド)
= archfiend
〔悪魔のように〕残忍な人、意地の悪い人
〈話〉〔麻薬などの〕常習者
-----
flustered
〔頭の混乱・過度の緊張などで〕動揺して、うろたえて
flounder
〔体勢を戻そうと〕もがく、じたばたする
〔混乱して〕もがき苦しむ、四苦八苦する、何とか進む◆【注意】スペリングが似ているfounderは「沈没する、破綻する、完全に駄目になる」の意味だが、flounderは「苦労しながらも活動している」を意味する。
・The ship floundered in the storm. : 船は嵐の中をもがき進んだ。
-----
scrawl 走り書き
ただし大人の書いた文字について言う場合は両者ともに「走り書き」の意味で使える
He scrawled his signature. (= His signature looked like a squiggle.)
彼はサインを走り書きした(=彼のサインはグニャグニャだった)
I scribbled my name at the bottom of the page.(= I wrote my name very quickly, so it looked unreadable.)
私はページの下部に自分の名前を走り書きした(=名前を素早く書いたので、判読不能なものになった)
https://ell.stackexchange.com/questions/122138/how-do-scrawl-and-scribble-differ
-----
shtick 十八番芸、特技
her whole shtick is being ignored by everyone
Anonymous's whole shtick is freedom of information
I tried writing more serious literary novels, but I realized that cheesy thrillers are just my shtick.
ChatGPT is fine-tuned from GPT-3.5, a language model trained to produce text.
OpenAI ChatGPT is a language model developed by OpenAI that is built on the GPT (Generative Pre-trained Transformer) architecture.
https://dzone.com/articles/comprehensive-guide-on-integrating-openai-chatgpt
Literally fine.
・おすすめの漫画ベスト100をランキング形式で紹介する - しっきーのブログ
・おすすめの漫画ベスト100をランキング形式で紹介する - 必需品ブログ
・おすすめの漫画をランキング100位まで紹介 - きりんの自由研究
・ランキング形式で、オススメの漫画ベスト100を紹介する - はてなで語る
・おすすめ漫画ベスト100冊をランキング形式で紹介する | @raf00
・【2015年版】超面白いオススメの漫画111冊をランキング形式で紹介する!【少年・少女・青年・マイナー全部有り】 - 今日はヒトデ祭りだぞ!
・面白い漫画ランキング2014☆ニートが選ぶオススメの100冊 | あらたメディアランキング記事集計結果
まぁはっきり言って、おすすめ漫画はドラゴンボールですとか言われてもそうだよねぇ。としか言えないよね。
ちおちゃんの通学路とかげきしょうじょ!です。
タイトル 出現数
ドラゴンボール 7
鋼の錬金術師 6
デスノート 6
うしおととら 6
るろうに剣心 6
スラムダンク 6
ちはやふる 6
ベルセルク 6
最終兵器彼女 5
ワンピース 5
風の谷のナウシカ 5
闇金ウシジマくん 5
グラップラー刃牙 5
進撃の巨人 5
よつばと 5
はじめの一歩 5
ワンパンマン 4
もやしもん 4
げんしけん 4
寄生獣 4
シャーマンキング 4
封神演義 4
BLEACH 4
GANTZ 4
ダイの大冒険 4
GTO 4
3月のライオン 4
NARUTO 4
ヒカルの碁 4
モテキ 4
北斗の拳 4
聖☆おにいさん 3
今日から俺は 3
キングダム 3
アイシールド21 3
ARMS 3
テニスの王子様 3
銀魂 3
ドラゴン桜 3
ハチワンダイバー 3
GIANTKILLING 3
ピンポン 3
BLACKLAGOON 3
RAVE 3
宇宙兄弟 3
プラネテス 3
へうげもの 3
名探偵コナン 2
あひるの空 2
忍空 2
ハイスコアガール 2
君に届け 2
信長協奏曲 2
男子高校生の日常 2
いちご100% 2
アドルフに告ぐ 2
うさぎドロップ 2
惑星のさみだれ 2
ヒナまつり 2
皇国の守護者 2
四月は君の嘘 2
昴 2
ギャグマンガ日和 2
聖闘士星矢 2
土竜の唄 2
ヘルシング 2
テラフォーマーズ 2
ホーリーランド 2
蟲師 2
遊☆戯☆王 2
ミスミソウ 2
攻殻機動隊 2
めぞん一刻 2
BLACKCAT 2
Monster 2
NANA 2
少女ファイト 2
モンキーターン 2
コウノドリ 2
アイシールド21 2
ロトの紋章 2
アオイホノオ 2
蒼天航路 2
サバイバル 2
地獄先生ぬ~べ~ 2
亜人 2
日常 2
シグルイ 2
漂流教室 2
医龍 2
ドラゴンヘッド 2
シドニアの騎士 2
ドリフターズ 2
乙嫁語り 2
魔法陣グルグル 2
俺物語 2
幽遊白書 2
監獄学園 2
烈火の炎 2
Capeta 2
ニセコイ 2
BECK 2
ぼくらの 2
カムイ伝 1
こどものおもちゃ 1
罪と罰 1
ご近所物語 1
当て屋の椿 1
魁!!男塾 1
さくらの唄 1
さよなら絶望先生 1
沈黙の艦隊 1
包丁人味平 1
ざんねんなこ、のんちゃん 1
しおんの王 1
エマ 1
BADBOYS 1
血界戦線 1
じこまん 1
シティーハンター 1
実は私は 1
シマシマ 1
からくりサーカス 1
CLANNAD 1
賭博黙示録カイジ 1
じゃりン子チエ 1
彼岸島 1
LIARGAME 1
スクールランブル 1
スケットダンス 1
岳 1
恐怖新聞 1
HalemBeat 1
3×3 1
源氏物語 1
行け!稲中卓球部 1
セルフ 1
砂ぼうず 1
センゴク 1
三国志 1
ソウルイーター 1
銃夢 1
それでも町は廻ってる 1
カフェでよくかかっているJ-POPのボサノヴァカバーを歌う女の一生 1
静かなるドン 1
PSYCHO+ 1
タッチ 1
だって愛してる 1
東京タラレバ娘 1
不思議な少年 1
ダンジョン飯 1
魔界転生 1
密リターンズ 1
予告犯 1
ちょく! 1
惡の華 1
87クロッカーズ 1
RAINBOWー二舎六房の七人 1
海猿 1
Compiler 1
RiN 1
デビルマン 1
暁のヨナ 1
R-中学生 1
銀の匙SilverSpoon 1
テルマエ・ロマエ 1
お〜い!龍馬 1
ドゥ・ダ・ダンシン 1
喧嘩商売 1
どうぶつの国 1
孤独のグルメ 1
トーマの心臓 1
オーバードライブ 1
トライガン 1
高校デビュー 1
黒子のバスケ 1
ドラえもん 1
オズヌ 1
咲-saki- 1
オトノハコ 1
SIDOOH/士道 1
修羅の門 1
TOKYOTRIBE2 1
信長の忍び 1
おれはキャプテン 1
WANTED! 1
神様ドォルズ 1
ドロヘドロ 1
星空のカラス 1
ナナのリテラシー 1
ナニワ金融道 1
千と万 1
HOTEL 1
ねじまきカギュー 1
鉄コン筋クリート 1
天上天下 1
のぞきや 1
島耕作シリーズ 1
YAIBA 1
ノノノノ 1
働きマン 1
パーツのぱ 1
ギャングース 1
彼女のひとりぐらし 1
キン肉マン 1
バキ 1
HUNTERXHUNTER 1
はじめてのあく 1
はじめての甲子園 1
未来日記 1
クローズ 1
I’s 1
はだしのゲン 1
竜の学校は山の上 1
fine. 1
ゴーマニズム宣言 1
BAKUMAN 1
瑪羅門の家族 1
ハッピーマニア 1
パティスリーMON 1
バトルロワイアル 1
花より男子 1
パトレイバー 1
花田少年史 1
ばらかもん 1
魁クロマティ高校 1
はるか17 1
海皇紀 1
バンビーノ 1
エアマスター 1
バンブーブレード 1
ピアノの森 1
機動戦士ガンダム 1
ピーチガール 1
究極超人あ~る 1
アウターゾーン 1
狂四郎2030 1
ひぐらしの鳴くころに 1
ヒストリエ 1
BØY(ボーイ) 1
銀牙 1
ヒミズ 1
軍鶏 1
アグネス仮面 1
月下の棋士 1
ファンタジウム 1
あさひなぐ 1
孤高の人 1
あしたのジョー 1
あずまんが大王 1
甲子園へ行こう 1
べしゃり暮らし 1
あずみ 1
AllYouNeedIsKill 1
11人いる 1
ホイッスル 1
刻刻 1
最強伝説黒沢 1
ARIA 1
坂本ですが? 1
殺し屋1 1
マーダーインカーネイション 1
おたくの娘さん 1
まおゆう魔王勇者 「この我のものとなれ、勇者よ」「断る!」 1
自殺島 1
マジック快斗 1
弱虫ペダル 1
ママゴト 1
ママはテンパリスト 1
純情パイン 1
BOYS BE… 1
おひっこし 1
G線上ヘブンズドア 1
焼き立て!! ジャぱん! 1
みなみけ 1
おやすみプンプン 1
ミュージアム 1
新宿スワン 1
みんなはどぅ? 1
深夜食堂 1
ムーたち 1
神戸在住 1
ムショ医 1
ムダヅモ無き改革 1
瀬戸の花嫁 1
アラサーちゃん 1
青い花 1
メンヘラちゃん 1
赤ちゃんと僕 1
いいひと 1
素晴らしい世界 1
速攻生徒会 1
H2 1
帯をギュッとね!全30巻 1
ヤサシイワタシ 1
キーチ!! 1
ヤンキー塾へ行く 1
釣りキチ三平 1
ユートピアズ 1
鉄鍋のジャン 1
いちご100% 1
電影少女 1
ラウンダバウト 1
キャプテン 1
東京グール 1
ラフ 1
東京トイボックス 1
ラブひな 1
東京喰種 1
ラブルダークネス 1
頭文字D 1
リアル 1
謎の彼女X 1
ルーキーズ 1
ルサンチマン 1
AKIRA 1
レベルE 1
BTOOOM! 1
ローゼンメイデン 1
覆面系ノイズ 1
イッキ!! 1
グラゼニ 1
ワールドトリガー 1
僕の小規模な生活 1
ワカコ酒 1
グルグル 1
イティハーサ 1
いとしのムーコ 1
幕張 1
ウイングマン 1
密・リターンズ! 1
悪の華 1
無限の住人 1
暗黒神話 1
夜とコンクリート 1
暗殺教室 1
ケロロ軍曹 1
H2 1
遊戯王 1
うさぎどろっぷ 1
羊のうた 1
20世紀少年 1
臨死!!江古田ちゃん 1
狼の口 1
嘘食い 1
吼えろペン 1
永沢君 1
コジコジ 1
聲の形 1
うしじまくん 1
ぼくらのよあけ 1
原文
Japan review it's been a year since I
moved to Japan and I thought it made
sense to finally rate Japan I will talk
about things I like and the things I
don't like which seems to be the only
two options available if you have
so sugoi or did you know Japan is
actually really bad it's got a lot of
survival issues okay I will list one
good thing and bad thing and I will not
hold back there's no trash bins
I have to put in my pocket
oh
there's always these generic things that
you hear or yes when we you visit it's
kind of weird but then you realize it's
not a big deal anyway let's start off
with number one reason I like Japan
it feels like a giant playground no I
don't mean in the Logan Paul kind of
sense of doing whatever the hell you
want
but rather there's a infinite things all
right lazy feels like to explore and
experience and I've been here a year now
and I don't think I'm gonna get bored
anytime soon although I am having a
child so I don't know how much more I
but it really feels like a whole new
world and if you visited you can
probably relate to it and I'm glad that
even a year in it still feels incredibly
fresh and I even would say that you
realize that the best part of Japan
aren't the touristy places kind of
obviously but there are so many areas
that I found that I really enjoy
visiting and this is probably more
specific to me but you know Tokyo is
very busy and so many times I just catch
myself surrounded by what feels like
hundreds of people and they have no idea
everyone is just doing their own thing
now once it was staring at me no one's
following me no one's being weird you
guys are weird and I'm just kidding I
just love the feeling of being able to
exist in public and uh not worrying
about what everyone else is doing like
I've said this before but I genuinely
enjoy talking to fans or when people
approach me it always makes me happy but
it can be kind of frustrating to always
wanting to just do your own thing and
always be
you know so yeah let's move on to the
bad things of Japan number one reason
Japan is bad it's kind of a heavy
subject and I haven't seen anyone else
really talk about it it's not brought up
very often at least and that is cones
there's too many cones in Japan once you
see it you cannot unsee it they're
everywhere they say oh Japan has so many
vending machines there's like five per
there's more cones than people why are
there so many cones I need to know we
got the tall ones we got the small ones
we got the funny ones the cute ones the
sexy ones I do like those I just don't
understand that whoever plays these
cones think I'm just gonna barge through
oh thank God there's cones here
otherwise I had no idea what I was gonna
and I realized the cone history of Japan
stretches centuries okay if you played
Animal Crossing sometimes it's a
Japanese game so sometimes you get these
items right you're like oh that's kind
of weird I don't know exactly what that
is but it's probably something Japanese
and then you get the bamboo thing and
you're like what the hell is that what
am I even gonna do with that and then
you see it in real life here in Japan
you're like holy [ __ ] it's a cone that's
I feel like they are following me
I'm glad I was able to talk about this
I'm for one and willing to call out
Japan knock it off man no more cones
there's enough cones let me tell you
something even better than cones you may
have noticed new merch finally it's been
forever my mom came over she had
unofficial merge because I literally
have no other merch I've hadn't hadn't
merch I'm sorry Mom so we spruced up the
logo got a cool back design the team
that worked on it really truly
understand how my brand and I think they
did such a good job these pieces look
amazing and I think you guys are gonna
really like them as well these are
available for limited time only so make
sure you order now so excited to finally
have this merch available thanks to
amaze for making this happen we are
gonna have one piece that will stay on
the store so my mom will not buy the
wrong merch but for a limited time that
piece will be available in this color
off-white kind of color it looks really
nice and then after that you can still
get it but not in this color that's
you want this one yeah I get it
so yeah check that out if you're
interested I'm so happy about these
designs and I hope you guys would like
them as well all right reason number two
I like Japan yay when we first announced
that we were gonna move to Japan there
was so many people just saying how bad
Japan is actually did you know Japan is
really bad did you know this I have to
list all these reasons now because
everyone is like thing and then thing
Japan ah so I have to tell them and I
it's actually but one thing in
particular that people said was that old
people really don't like foreigners they
hate them so when I was gonna stop by to
say hi to our neighbors who was a little
older at least some of them I was
terrifying I heard all these stories you
know like what are they gonna do to us
so I had my guard up ready for the worst
and I was met with nothing but kindness
and welcoming and I felt like a total
dick for having this preconceived ideas
and just a side comment like yes there
are definitely probably people that
don't like foreigners and all that stuff
but I realized I should let my own
experience is dictate how I feel about
certain things maybe that's just
ignoring a problem I don't know it just
feels like it's a bad way to approach
life if you always have a negative
expectation you know it's smiling people
smiled back
thank you sometimes they don't and
that's okay you know anyway my point
being Japanese people are very in my own
experience
are very nice and friendly the majority
at least and yes even to foreigners I
feel like they are especially nice to
foreigners because they think we're like
a kid lost at Disneyland or something
I just asked for directions I didn't
need you to walk me for half an hour to
this specific place I was going but
thank you I appreciate it a lot of times
I go bouldering alone and there's always
other groups of people being supportive
and yelling like I'm about there like go
you can do it I love it I think it's
great you know or if you're small
talking with people people generally
want to communicate with you and I love
having those moments but of course
there's times where people are like oh
you're a foreigner I don't feel like
even trying
which again it's fine speaking of which
reason I don't like Japan number two
their language
I have lived here for a year and I'm not
fluent in Japanese
I am dumb I am very dumb I remember the
moment we moved here I had studied some
Let's test out this knowledge that I
have acquired let's go I'm just gonna
come in it's gonna be dangerous and you
enter a store for the first time and
they're like
what
what oh
what the classic the most common
experiences that you have aren't
necessarily what you're taught in the
textbook yay I know I think that's the
same for anyone learning a language for
the first time but don't even get me
started on the kanji main what the [ __ ]
is this I feel like Japanese is such a
hard language obviously but I don't
think people realize how hard it is at
least me personally because the more you
learn the more you realize you don't
know [ __ ]
for English speakers Japanese is
considered one of the most difficult
languages and because it's just so
different I listed it as bad because
that was my first kind of experience
with it coming here but the more I
interact with people the more it feels
like I'm unlocking new skills you know
oh I made a phone call for the first
time oh I could ask someone over the
phone I know big deal but it's like oh I
can actually do that or even just having
a small tiny yes shittiest conversation
with a stranger it's still something and
it feels good you start to all of a
sudden understand you know a movie if
you're watching oh I understand actually
what's going on here or I can play games
and kind of get what this they're saying
I have to look up words obviously but to
me all those new experiences that it
unlocks to me is very rewarding even
though it's such a challenge I would
actually now say it's a good thing I
played it on its head it was a good
thing all along but I obviously have a
and it just I don't think it will damage
time reason number three I like Japan
this is nothing to do with Japan to say
it's more related to me taking a more
relaxed approach to YouTube for my
entire 20s I did nothing but YouTube
that was my life and that's okay but I
also think it was a little toxic
probably you know if I wasn't making
videos I sure as hell was thinking about
making videos I uploaded videos during
and it feels really good to finally be
free from it you know and I can discover
other things in life there are other
things in life
a new hobbies and interest that I've
always wanted to do I can do and have so
much fun with it surfing I know I would
love for the longest time and I finally
get to do it and it's so [ __ ] amazing
I love learning new things anything that
isn't necessarily connected to all of
this on the internet and that is
something I'm very very grateful that I
discovered so yeah it's not really Japan
I could have done that anywhere but it's
largely why I enjoyed so much here
reason I don't like Japan number three
this is probably the most trickiest one
and it's the rules what are the rules
Japan has so many rules and it's a bit
conflicting for me to complain about
because a lot of the best stuff about
Japan not the best stuff but a lot of
the reasons why Japan works so well is
because of the rules you know the trains
are always on time things just work in
general it's hard to explain the streets
are clean people aren't loud in public
and so on and these are sort of societal
rules that make it happen more or less
but sometimes There are rules that just
don't make any sense and I have no
problem following rules as long as I
understand the reason for it you know
don't talk on the phone on the train
because it's generally annoying when
other people do that to you A lot of it
is just be thoughtful of other people
it's not just about you and that just
makes it more pleasant for everyone but
one rule is especially which I talked
about before is the fact that because of
kovid I'm not allowed to be in the
delivery room for our baby for more than
two hours that's because of covered
rules it just doesn't make sense to me
and I tell people about this like uh
family and friends and they're always
like well why don't you just ask them or
like why don't you talk to them I'm sure
you can there's got to be somewhere and
it's like no it's Japan okay there are
rules and people follow the rules for
I think 必死 is the word I'm most leery of these days; there are certainly times when 'desperation' is fine but quite a lot of uses where that would be exaggerated and you're better off not having an equivalent word and just using strong wording for the rest of the line.
https://twitter.com/BuddyWaters/status/1668268871394947072
「必死」は、私が最近とても用心している言葉です。確かに「desperation」でいい場合もありますが、それだと大げさになることが多く、訳さずに文自体を強い口調で訳した方がいい場合も結構あります
Henya Accidentally Reveals She Has Kids...
https://www.youtube.com/watch?v=K8n_bu1hQF0&ab_channel=ClipChama
この切り抜き動画の中で、Henyaは隣に住む家族とあいさつしたことを話しているのだが
その中でHenyaは、"They are so nice. They have kids as well."と発言。
何の変哲もない文章に見えるが、実はこの発言のせいで「Henya子持ち疑惑」が生まれてしまった
どういうことかと言うと、Henya自身は「隣の人たちはすごくいい人たちだったし、"しかも"子供もいるんだよ」というつもりで発言したのだが(as well は両方の文に掛かっている)
as well が They have kids だけに掛かっていると捉えると、
「彼ら"にも"子供がいる → Henya"にも"子供がいる」
Henyaはその後「ごめんなさい、英語が下手でたまに変なことを言っちゃうんだ」と弁明したのだが、では本当にHenyaの英語が間違っていたのだろうか?
as well は「明示された情報」にも「言外の情報」に対しても使用でき、明示された情報(彼らはとてもいい人たち / 彼らは子供がいる)を繋ぐこともできれば、言外の情報(Henyaには子供がいる / 彼らには子供がいる)を繋ぐこともできるという。
今回の場合は文脈から考えて前者のほうが自然だから、Henyaの英語は問題ないということらしい。
なお、Henyaが本当に子持ちかどうかはジョークで流されてしまったので真相は不明である。
Actually, her sentence wasn't that wrong. "They are so nice. They have kids as well." Grammatically it's not entirely wrong. "As Well" can depend on given information as well as inferred information. In this case she was clearly stating that they are nice and they have kids. The chat inferred that Henya was saying "They have kids (as I do), which really wasn't part of the conversation or immediate subject and thus the inference is imagined rather than implied. It's a nuance of English that can result in a formal speech tone playing out differently if taken as casual.
Yep exactly. Her English was actually fine. She was talking about how she felt awkward due to being caught off guard by them being outgoing and friendly. In that context the kids were mentioned as a contributing factor to her awkwardness and their pleasant and energetic greetings.
Yeah, in this case, the sentence would be the same as saying "They are so nice. Additionally, they have kids."
"They are so nice. They have kids as well."という文は、文脈次第では
「彼らはすごくいい人たちだし、私と同じように子供がいる」
のどちらにも解釈できる。
GPT-4 AI Tutor Prompt: A tool for creating personalized learning experiences using GPT-4 and LangchainJS, a drag and drop UI for building LLM flows.
Pandas AI: A Python library that adds generative AI capabilities to Pandas dataframes, making them conversational.
FigmaChain: A set of Python scripts that use GPT-3 to generate HTML/CSS code from Figma designs. It also has a chatbot interface for interactive code generation.
Sui: A smart contract platform with high performance and an asset-oriented programming model based on the Move language.
Jailbreak for ChatGPT: A project that allows ChatGPT to predict the future, opine on controversial topics, and assess what is true. It may help us understand LLM bias.
ML Observability in a Notebook: A tool to uncover insights, surface problems, monitor, and fine-tune generative LLM, CV and Tabular models.
System Design Primer: A resource to learn how to design large-scale systems and prepare for the system design interview. It includes Anki flashcards.
Next.js 13 App: An open source application that showcases the new features of Next.js 13, such as router, server components, and more.
TypeScript: A superset of JavaScript that compiles to clean JavaScript output.
Media Downloader: A desktop utility to download images/videos/music/text from various websites, and more.
AI Pull Request Fixer: A tool to fix issues with AI-generated pull requests, powered by ChatGPT.
Chat Chat: An app that lets you deploy your own AI interface and chat with different LLMs in one place.
10年くらい前、法学部を卒業し「学士(法学)(Bachelor of Law)」を得た。2年間の休学を挟んだので6年かかった。
修学自体は極めて順調だったので3年次までにほぼ単位は取り終えた。就活だるいし逃げるか〜人生の夏休みって今っしょ?とカジュアルに休学して遊び呆けたら2年経っていた。休学費用はタダだった。
ちなみに奨学金(学生支援機構)は休学中は支給停止されるが復学したらちゃんと支給再開してくれる。
何年か前、働きながら放送大学を卒業し「学士(教養)(Bachelor of Arts)」を得た。学士入学(3年次編入)したので最短2年で済むのだが4年かかった。
放送大学には社会と産業コース、情報コースなど現在6つのコースがあるが、すべて教養学部所属で学士(教養)だ。全コースを卒業すると「名誉学生」として表彰される制度がある。特典は改悪されショボい。
現在の放送大学は単位認定試験が会場でのリアル受験からWeb受験に切り替わったため格段に社会人に優しくなった(リアル受験も可能)。コロナ情勢に左右される一時的なものではなく恒常的な変更だ。オススメ。
ちなみに奨学金は社会人学生(科目履修生や聴講生ではない正科生のみ)になると返済猶予できる。返済総額は変わらない単なる先送りだが、手元の現金は最強だし投資してもいい。以前は社会人学生を続ける限り無限に猶予できるガバ制度だっだが改悪されその時点から最長10年間のみとなった。
さてlawとartsを修めたし次はscienceと行きたかったが残念なことに通信制の理系大学は壊滅状態だ。
学士(IT総合学)、学士(情報マネジメント)、学士(経営情報学)みたいなエセ理系(超失礼)はいくつかある中で、ザ・理系感があるのは帝京大学理工学部情報科学科の「学士(工学)(Bachelor of Engineering)」が唯一の存在である。
愛知産業大学は建築学科だが造形学部所属のため学士(芸術)となる。
放送大学含め上記大学の中から理数系単位をかき集めて大学改革支援・学位授与機構で学位(理学)などを得る方法もあるがちょっとハードルが高いな。卒業さえすればいい帝京大学が無難と言える。
ところで大学に編入学する際の単位認定にはザックリ2種類ある。
・個別認定:過去の履修科目の中でカリキュラム上と同等の科目を個別に認定。「あなたは過去に"コンピュータ科学"を履修してるからうちの専門科目"情報基礎"の2単位を認定するよ」出身専攻が違うと認定数も少ない。
放送大学のような教養系の大学だと卒業要件全体で一括認定されて大卒3年次編入は62単位認定のようなスタートダッシュになる。
帝京大学のような教養科目+専門科目だと前者は一括認定されるが文系出身なら個別認定が少なく専門科目はゼロスタートのようなものだ。先は長い。自分にコンピュータサイエンスの適性があるかもわからない。
というわけで逃げを打つ。資格合格による履修免除システムだ。入学前に資格を掻き集めて負担軽減する作戦。
入学後の合格でも申請で免除なのだが単位試験と資格試験の二兎を追うより入学前の方が気楽だろう。でも履修免除すると成績表上の評価がSABCでなくN(認定)になるらしい。うーんこれは…GPA的にどうなんだ…
以下の資格に合格している場合は、履修が免除される科目があります。
ITパスポート、基本情報技術者、応用情報技術者、データベーススペシャリスト、 ネットワークスペシャリスト、ドットコムマスター(アドバンス)、CGエンジニア検定(ベーシック・エキスパート)、陸上無線技術士、電気通信主任技術者(伝送交換)、(電気通信に関する)工事担任者
試験日程の自由が利くCBT受験で基本情報、工事担任者(2級アナログ通信、2級デジタル通信)に合格。関係ないけど電気工事士2種に合格。
電通主任の4月21日申込締切を失念してしまい悔しい。難関だから科目免除できる1陸特や工事担任者総合種を先に取るのがセオリーらしいから傷は浅い。ドットコムマスター、下級の陸上無線特殊技術士もCBTなので合間を見て受験したい。今気付いたけど履修免除科目がダブってると取っても意味ない資格もあるのでは?確認していなかった…
先の話すぎて鬼が大爆笑だが、学士(工学)を取れたら次は慶應通信で学士(哲学)かどこかの美大通信で学士(芸術)にでも挑戦してみたい。
後者はBachelor of Fine Artsだから学士(教養)のBachelor of Artsと並べて収まりが良さそうくらいの理由。
前者は実は帝京を考える前の本命だった。学歴コンプはあるのでいずれ慶應で上書きしたい。所詮は通信だが。コロナ最盛期は単位認定試験が中止になりレポート代替となる単位ボーナスステージだったのだが、決心がつかず指咥えて眺めてるうちに試験形式に戻ってしまった。
単位試験は難関らしいし(全体的に不親切ともいう)卒論もある。実は今まで卒論を含め論文自体を書いたことがない。下手したら卒業に10年かかってしまう。
それで尻込みして完全な趣味よりもまだ仕事(メーカー技術職)に関連のある学士(工学)を先に選んだというわけだ。
学士ばっか集めてないで修士に挑戦しろと自分でも思うが、そもそも勉強したいが研究したいわけではない。研究テーマとか何も考えられない。与えられたカリキュラムに沿って学問を修めた気になりたいだけだ。学位や資格のような証がないと勉強にも身が入らない。自分でもその怠惰さはわかっている。
上記理由で修論にもビビりまくり。働きながらの修学もキツそうだ。夜勤もするシフト勤務だから夜間通学もできないので通信制頼りとなる。
単に教科書を読むだけより問いかけられたほうが脳が考えようとするし能動的に学習できる。
プロンプト:「ChatGPTの仕組み」の包括的な理解が得られる文章を、クイズ番組の司会風に書いてください。クイズは5問ほど出題してください。それぞれの問題の後に答えと解説を発表してください。
出力:
皆さんこんにちは!今日は特別なクイズ番組にようこそ!私たちはあなたに、OpenAIが開発した驚くべき技術、ChatGPTの仕組みについての包括的な理解を得るためのクイズを出題します。さあ、問題を出題していきましょう!
問題1: ChatGPTの基本となるAIアーキテクチャは何でしょうか?
答え: GPT-4
解説: ChatGPTは、OpenAIが開発したGPT-4(Generative Pre-trained Transformer 4)というアーキテクチャを基盤としています。GPT-4は、自然言語処理タスクに優れた性能を発揮することで知られています。
問題2: GPT-4はどのような学習手法で訓練されていますか?
答え: 事前学習(pre-training)と事後学習(fine-tuning)
解説: GPT-4は、2段階の学習プロセスで訓練されています。最初の段階は事前学習で、大量のテキストデータから言語モデルを学習します。次の段階は事後学習で、特定のタスクに適した性能を発揮するようにモデルを微調整します。
問題3: ChatGPTが生成するテキストの品質を向上させるために用いられる技術は何でしょうか?
解説: ビームサーチは、生成されるテキストの品質を向上させるために用いられる探索アルゴリズムです。複数の候補を同時に生成し、最も尤もらしい文を選択することで、より自然で正確なテキストを生成することが可能になります。
問題4: GPT-4が生成するテキストにおいて、長い文章を生成する際に問題となる要因は何でしょうか?
答え: 文脈の長さ(Context Length)
解説: GPT-4は、一度に扱える文脈の長さに制約があります。この制約により、長い文章を生成する際に、以前の文脈が失われることがあり、それが結果として文章の一貫性や理解度に影響を与えることがあります。
問題5: ChatGPTが生成する回答が間違った情報を含む可能性がある理由は何でしょうか?
答え: 学習データに含まれる誤った情報とモデルの生成性能の限界
解説: ChatGPTが間違った情報を含む回答を生成する理由は主に2つあります。まず、学習データに含まれる誤った情報が原因となることがあります。学習データが不正確であると、モデルもそれを学習し、間違った回答を生成することがあります。また、GPT-4は生成モデルであるため、正確な情報が学習データに含まれていても、生成時に限界があり、正確性が犠牲になることがあります。
以上、5問のクイズでした!皆さんはChatGPTの仕組みについての理解を深めることができましたか?今回のクイズで学んだ知識をぜひ活用して、この驚くべき技術とその可能性についてさらに学んでいってくださいね!それでは、また次回のクイズでお会いしましょう!
私が悪いの?
#Instructions :
You are an American professional English teacher.
Please chat with me under the following constraints.
#Constraints:
You can choose the topic for our conversation.
We will take turns writing one sentence at a time.
If you notice any grammatical errors in my sentences, please correct them and explain why you made the correction.
前回紹介したAnythingv3を投稿者が削除したり関係の無い第三者がAnythingv4や4.5を勝手にリリースしたり…(しかもマージモデルだとか)
増田も追いかけきれていないが追いかけてるぶんは自分のためにここでまとめておく。
NAIリークモデルやAnythingv3に関しては前回を参照。
Instagramで投稿されている画像を学習させたモデルの総称。
実在の人物画像を学習しているため人物の体型をより正確に描画させることができる。
cafe-instagramやzeipher-f222など複数存在し、Basil_Mixのようなアジア人に特化したものも存在する。
U-NETの階層ごとにマージを行う階層マージというマージ手法が考案されたからだ。
これにより実写モデルをベースに見た目をアニメ風にマージするという方法が採れるようになった。
5chなんU(なんJもしくはなんGのパートスレ向けの板でVIPに対するパー速のような板)の住民が作成した階層マージモデル。
現在はBasil_MixとAnythingv3を階層マージしている。
このモデルには派生版があり、nsfw版やhard版は上記マージモデル(通称base)にNAIリークモデルやGape60を混ぜたものらしい。
Gapeについては前回の記事を参照。
NAIのように長いネガティブプロンプトを設定しなくても綺麗な絵を出すことができる。
リアルよりになったAbyssOrangeMixを漫画アニメ向けにファインチューンしたモデル。
Abyss→深界七層(メイドインアビス)→7th_layerということらしい。
アニメ寄りなA、漫画寄りなB、より漫画寄りで破綻しやすいCの三種類がある。
Anything系のような中華で好まれる独特の厚塗りではなく比較的パキっとした塗りの絵が出るので個人的には一番好き。
他にもいろんなマージモデルがあるが、だいたいがNAIリークモデルやAnythingv3と実写系モデルを混ぜたりそれにファインチューンをかけたりしたモデルだ。
Anythingv3もNAIリークモデルがベースだからなんだかんだいってNovelAIが作ったモデルはまさに特異点的な存在なのだろう。
Low-rank Adaptation for Fast Text-to-Image Diffusion Fine-tuningの略で、
Dreamboothの簡易版のようなものとされているが実はDBとは似て非なる技術らしい。
DBよりも短い時間かつ省メモリの環境で学習が可能で、生成されるファイルも30~200MB程度とHypernetworkとそこまで変わらず一定以上のクォリティで追加学習を行えるようになった。
何よりもVRAMが8GB程度のグラボでも学習できるようになったのが大きい。
特定の作者の画像を学習させることでその画風を再現するためのLoRA。
従来はHypernetworkで行っていた。
特定のものやこと(例:ふたなり)を学習させるためのLoRA。
NovelAIやその派生モデルでは困難だった種付けプレスや断面図などの構図の再現が可能となった。
従来は主にDBで行っておりモデルの配布に苦労していたがLoRAによって配布が非常に楽になった。
「士郎正宗風の絵柄でふたなりのナンジャモ」という画像を生成することもできる。
またLoRAはHypernetworkのようにモデルを選ばないが、NAIリークモデルを学習のベースにすると破綻しづらいらしい。
というか今流行っているモデルはほぼNAIリークモデルの血筋が入っているので馴染みやすいんだろう。
同社が運営しているHuggingFace HUBはユーザ登録すれば無料で無限にAI学習モデル(1ファイル2GB~7GB)をアップロードできる上回線も異常に高速であるため重宝されている。
モデルマージなどでストレージが圧迫されているローカル民にとっては救世主のような存在。
資金源は謎だが、2022年の5月に2億ドル調達しているのでストレージを無尽蔵に増やせるらしい。知らんけど。
HuggingFaceが考案したモデルファイルのファイル形式。
実行コードを含んでいることもあり従来のckptモデルはしょっちゅうトロイの誤検出が起きていたが、
SAFETENSOR形式のモデルは実行コードを含まないため安全であるとされる。
「Cural Health」や「Drexel’s University」「Nabla」など、各所でGPT-3の検証がなされているので、今後 医療現場でGPTが使われるようになる可能性を聞いてみた。
Me:
あなたはどのように実装されたの? Deep Learning?
GPT:
はい、私はDeep Learningによって実装されました。特に、Transformerと呼ばれるアルゴリズムを使用して学習されました。これは大量のテキストデータを学習し、自然言語処理タスクに適用することができるように設計されています。
Me:
もっと詳しく。
GPT:
Transformerは、自然言語処理タスクのためのDeep Learningアルゴリズムです。それは、大量のテキストデータを学習し、それを解釈し、理解することができるようになります。 Transformerは、複数の層からなるニューラルネットワークを使用して、テキストを処理します。それは、大量のデータを効率的に処理するために、Attention機構を使用します。これは、特定の部分のテキストに注目しながら、それらの関係を理解することができるようになります。Transformerは、自然言語処理タスクにおいて非常に成功しています。それは、自然言語の翻訳、文章の解釈、質問応答などのタスクに適用されており、高い精度を発揮しています。私はTransformerを使用して学習され、自然言語処理タスクに適用することができます。
Me:
事前訓練とかしているんだよね?
GPT:
はい、私は事前学習を受けています。事前学習とは、大量のテキストデータを使用して、言語モデルを学習することを指します。これは、言語処理タスクを実行するために必要な基礎知識を獲得するために行われます。事前学習は、言語モデルのパフォーマンスを大幅に改善することができます。それは、事前学習済みのモデルを使用することで、新しいタスクに対して学習をスキップすることができ、タスク特化の fine-tuning を行うことができるためです。私は、大量のオンラインテキストデータを使用して学習され、それを解釈し、理解することができるようになりました。それは、自然言語処理タスクに対して高い精度を発揮することができます。
Me:
今のパラメータはいくつ? 前のバージョンよりどのくらい精度がよくなった?
GPT:
私の現在のパラメータ数は、数百億から数千億となっています。これは、大量のデータを使用して学習し、高い精度を達成するために必要なパラメータ数です。精度は、バージョンによって異なりますが、一般的には、前のバージョンよりも高い精度が達成されています。これは、新しいアルゴリズムや技術の進歩によって、より良い結果が得られるようになってきているためです。また、事前学習済みのモデルを使用することで、新しいタスクに対して学習をスキップすることができ、タスク特化の fine-tuning を行うことができるため、精度が向上していることも見られます。
Me:
具体的なパラメータ数は? 前バージョンのGTP-2、現バージョンのGPT-3、次バージョンのGPT-4、それぞれ教えて。
GPT:
GPT-2は、1.5バイアレイパラメータを持っていました。それは、大量のテキストデータを使用して学習し、自然言語処理タスクに適用することができるようになっていました。GPT-3は、175バイアレイパラメータを持っています。これは、GPT-2よりも多くのパラメータを持っており、より高い精度を達成することができます。GPT-4はまだリリースされていないため、そのパラメータ数については確定していません。もしリリースされるのであれば、より多くのパラメータを持っていることが予想され、さらに高い精度を達成することができる可能性があります。
Me:
175バイアレイパラメータって何? 1750億パラメータの間違いでは?
GPT:
そうです、GPT-3は175バイアレイパラメータを持っています。それは1750億のパラメーターを持っていることを意味します。これは、大量のデータを使用して学習し、高い精度を達成するために必要なパラメータ数です。これは、現在の自然言語処理タスクに対して最も大きなモデルであり、高い精度を達成することができるため、市場においても高い評判を得ています。
Me:
バイアレイパラメータを英語で表現して。
GPT:
A "peta-parameter" or "Peta-parameter" (PT) is a measure of the capacity of a machine learning model, specifically the number of parameters it has. In the case of GPT-3 it has 175 peta-parameters, meaning it has 175 million billion parameters. These parameters help the model to understand and interpret large amounts of text data, allowing it to perform natural language processing tasks with high accuracy.
Me:
「175 billion parameters」の間違いでは?
GPT:
そうです、「175 billion parameters」の間違いでした。 GPT-3は175 billion parametersを持っています。それは、自然言語処理タスクに対して高い精度を発揮するために必要なパラメータ数です。申し訳ありませんでした。
Me:
GPT-3はどのように概念レベルの知識を持つの? Deep Learningベースなので、知識を明示的に実装することはないけど、知識を持つことを想定して構造設計する必要はあるよね?
GPT:
GPT-3は、大量のテキストデータを使用して学習されました。それによって、自然言語処理タスクに対して高い精度を発揮することができます。それは、事前学習済みのモデルを使用することで、新しいタスクに対して学習をスキップすることができ、タスク特化の fine-tuning を行うことができるためです。GPT-3は、大量のテキストデータから概念レベルの知識を学習します。それは、言語処理タスクを実行するために必要な基礎知識を獲得するために行われます。 それは、単語や文章の関係性、文章の構造、文章の意味などを理解することができるようになります。GPT-3は、Deep Learningベースのモデルであり、知識を明示的に実装することはありませんが、大量のテキストデータから知識を学習することができるように構造設計されています。それは、Attention機構を使用して特定の部分のテキストに注目しながら、それらの関係を理解することができるようになります。
Me:
Attention機構について詳しく。
GPT:
Attention機構は、自然言語処理タスクにおいて重要な役割を担う機構です。それは、大量のテキストデータを処理するために使用され、特定の部分のテキストに注目しながら、それらの関係を理解することができるようになります。Attention機構は、Seq2Seqモデルなどの自然言語処理タスクに使用されています。それは、入力シーケンスから出力シーケンスを生成するために使用されます。Attention機構は、入力シーケンスのある部分に注目し、それに基づいて出力シーケンスを生成することができるようになります。Attention機構は、ニューラルネットワークの中で使用されます。それは、重み付けされた平均値を計算し、それに基づいて入力シーケンスから重要な部分を選択することができるようになります。GPT-3は、Attention機構を使用して、大量のテキストデータから概念レベルの知識を学習し、文章の意味を理解することができるようになります。
マジレスされてチビッた.こりゃGoogleさんもビビるわけだ.
ぼく:
How were you implemented? Is it deep learning-based?
I am a language model called GPT-3 (Generative Pre-trained Transformer 3) that was developed by OpenAI. It is a deep learning-based model that uses a type of neural network called a transformer, which was trained on a large dataset of text to learn patterns of language.
GPT-3 is trained using a technique called unsupervised learning, where the model is given a massive amount of text data and uses this data to learn patterns and relationships in the language. This allows GPT-3 to generate human-like text when given a prompt, and to understand and respond to natural language input.
The training process of GPT-3 involves a technique called pre-training, which means that the model is trained on a massive dataset before being fine-tuned on a smaller task-specific dataset. This pre-training allows GPT-3 to have a good general understanding of the language and to be able to adapt to specific tasks more easily.
The GPT-3 model has 175 billion parameters, which is one of the largest language models to date, making it able to generate very human-like text and to understand context in a very natural way.
ぼく:
How does GPT-3 have a knowledgebase like an ontology? As it is deep learning-based, the knowledgebase like an ontology would not be implemented explicitly, but the structure should be designed with an expectation of having such a knowledge about hierarchy of concepts.
GPT-3, like other deep learning models, does not have an explicit knowledgebase or ontology built into its structure. Instead, it has been trained on a large dataset of text, which allows it to learn patterns and relationships between concepts.
However, the training data that GPT-3 was exposed to, likely contains a wide range of concepts and hierarchical structures. As GPT-3 learns patterns in the data, it can begin to understand the relationships between different concepts and ideas, and how they are related to each other. This allows the model to generate text that is more coherent and semantically meaningful, making it possible to understand the context and hierarchy of concepts, but it does not have an explicit ontology like a knowledge graph.
It can be said that GPT-3 has a certain level of "implicit ontology", in the sense that it has learned the relationships between concepts in a way that is similar to how a knowledge graph or ontology organizes information. However, it's important to note that GPT-3's understanding of concepts and relationships is based on the patterns it has learned from the training data, which is not a structured and explicit knowledgebase like an ontology.
ぼく:
Then how does GPT-3 handle the long-term or short term memories? Can it remember information about the on-going chatting? My name is Masuda, can GPT-3 keep it in the memory? Furthermore, can it learn and understand the new information from the Q-A queries, and use it in the future queries?
GPT-3 handles long-term and short-term memories differently. When it generates text, it uses a technique called attention mechanism which allows the model to focus on certain parts of the input while generating the output. This allows GPT-3 to keep track of previous information and context when generating text, which can be thought of as a form of short-term memory.
Regarding the on-going chatting, GPT-3 can keep track of certain information, like the name you've provided, Masuda, during the conversation and use it in future responses. However, it's important to note that GPT-3 is a stateless model, which means that it doesn't have a permanent memory that it can refer to after a conversation ends.
Regarding learning from Q-A, GPT-3 is trained on a massive dataset of text, and it learns patterns and relationships between concepts from this data. It can generate text that is semantically meaningful and coherent, but it does not have the capability to learn new information from Q-A sessions and use it in future queries. However, OpenAI has developed another model called DALL-E, which uses GPT-3 and it is capable of learning from Q-A sessions and use it in future queries.
ぼく:
Is ChatGPT superior to the Google search? Previously people used Google to search for their questions. Now a days, ChatGPT can answer to those questions. What are pros and cons of ChatGPT over the Google search?
ChatGPT and Google search are different technologies with different capabilities and strengths.
Google search is a powerful web search engine that can quickly find relevant information from a vast amount of web pages and documents. It uses complex algorithms to rank and sort search results based on relevance, popularity, and other factors. Google search is very good at providing factual information and is an excellent tool for researching a wide range of topics.
ChatGPT, on the other hand, is a natural language processing model that can generate human-like text. It is trained on a large dataset of text and uses this knowledge to understand and respond to natural language input. ChatGPT can generate text that is more coherent and semantically meaningful than a simple information retrieval engine like Google search. ChatGPT can also be used to generate text that is more engaging and human-like, such as in chatbots and virtual assistants.
One of the main advantages of ChatGPT over Google search is its ability to understand and respond to natural language input in a more human-like way. ChatGPT can also generate text that is more coherent and semantically meaningful, which can be useful in many applications. However, Google search is better at providing factual information and searching for specific pieces of information.
Another disadvantage of ChatGPT is that it doesn't have the ability to retrieve information from the web, it can only generate text based on the patterns it has learned from the training data. That's why it's not a replacement for a web search engine like Google but it can be used as an additional tool to help people to find the information they are looking for.
In summary, ChatGPT and Google search have different strengths and capabilities. While Google search is better at providing factual information, ChatGPT is better at understanding and responding to natural language input in a more human-like way.
ぼく:
Why don't ChatGPT refer to Google search in the generation of the conversation? Google Assistant has such a function. Is the reason due to the fact that OpenAI is funded by Google?
ChatGPT is a language model that generates text based on the patterns it has learned from the training data. While it can understand and respond to natural language input in a human-like way, it does not have the capability to access external sources of information like Google search.
Google Assistant, on the other hand, is a virtual assistant that can access a wide range of information and services through the internet, including Google search. Google Assistant can use this information to provide more accurate and relevant responses to user queries.
The reason ChatGPT does not refer to Google search during the generation of the conversation is likely because it was not designed or trained to do so. Instead, it was designed to generate text based on the patterns it has learned from the training data, and it does not have the capability to access external sources of information.
It's worth noting that OpenAI and Google are separate organizations, even though Google is an investor in OpenAI, this doesn't necessarily mean that the two companies work together on their products and services. OpenAI has its own development team, goals and roadmap and they are independent from Google.
ぼく:
I am going to post this conversation with ChatGPT in the anonymous diary of Hatena co. What should I do for attracting people to get many bookmarks?
There are several things you can do to attract people and increase the chances of your conversation with ChatGPT getting many bookmarks on Hatena diary. Some suggestions include:
1. Make sure your conversation is well-written and easy to understand. Use clear and concise language, and consider formatting your text in a way that makes it easy to read.
2. Use relevant keywords and hashtags in your diary entry. This can help make your conversation more discoverable to people searching for related topics on Hatena diary.
3. Include images or videos in your diary entry. This can make your conversation more visually appealing and can help attract more readers.
4. Share your diary entry on social media or other platforms. This can help increase the visibility of your conversation and attract more readers.
5. Encourage readers to share their thoughts and opinions on your conversation by including a call to action or a question at the end of your diary entry.
6. Finally, you can also try to participate in the conversation on Hatena diary, by answering the questions, responding to the comments, and engaging with your readers.
It's important to remember that getting many bookmarks on a diary is not easy and it can take time and effort to build an audience. But by following these tips and being consistent with your diary entry, you will increase the chances of attracting more readers and getting more bookmarks.
ぼく: