米国を中心とした 人工知能(AI)の事業トレンドを紹介する.“AI”というと私がリードしたドコモの 「しゃべって コンシェル」やApple Siri ,Amazon Echo, Microsoft Cortanaなどの擬人化エージェント,あるいはロボットを想起することが多いが,米国ではより一般的にAIは農業や交通,流通など各産業における自動化,効率化技術のマーケティング用語として用いられる場合が多い.実質はデータに基づく機械学習を用いた最適化である.基盤となる実装アルゴリズムの多くは急速にコモンディティ化して行く.画像解析,パーソナル化,音声対話,マーケティングオートメーションの多くは商用化段階にありレッドオーシャン化が進みつつある.農業,教育,医療,法律,福祉,防災などこれまでIT化が進んでいなかった領域での動きを含めてAIの事業化動向を解説する.
Future Standard では、IoTを活用した映像解析のサービス開発を行っており、その仕組みを支える技術についてお話をさせて頂きます。前半では、AWSでサーバー側の仕組みを構築した際に、どのような設計思想に基づいてアーキテクチャを構築したのかといったポイントを中心に説明します。後半は、NVIDIAのJetsonという組み込みコンピューターを使った、Faster R-CNN を使ったエッジでのリアルタイム物体認識の取り組みについてお話させて頂きます。
米国を中心とした 人工知能(AI)の事業トレンドを紹介する.“AI”というと私がリードしたドコモの 「しゃべって コンシェル」やApple Siri ,Amazon Echo, Microsoft Cortanaなどの擬人化エージェント,あるいはロボットを想起することが多いが,米国ではより一般的にAIは農業や交通,流通など各産業における自動化,効率化技術のマーケティング用語として用いられる場合が多い.実質はデータに基づく機械学習を用いた最適化である.基盤となる実装アルゴリズムの多くは急速にコモンディティ化して行く.画像解析,パーソナル化,音声対話,マーケティングオートメーションの多くは商用化段階にありレッドオーシャン化が進みつつある.農業,教育,医療,法律,福祉,防災などこれまでIT化が進んでいなかった領域での動きを含めてAIの事業化動向を解説する.
Future Standard では、IoTを活用した映像解析のサービス開発を行っており、その仕組みを支える技術についてお話をさせて頂きます。前半では、AWSでサーバー側の仕組みを構築した際に、どのような設計思想に基づいてアーキテクチャを構築したのかといったポイントを中心に説明します。後半は、NVIDIAのJetsonという組み込みコンピューターを使った、Faster R-CNN を使ったエッジでのリアルタイム物体認識の取り組みについてお話させて頂きます。
日本ではディープテック・スタートアップが育つ環境がない。1. イノベーター人材育成が不十分、2.ベンチャーキャピタル側の人材欠如、3. 大企業側レセプターが未発達という課題を指摘した上で、今後どうすれば良いのかという未来志向の議論をしたい。巷間で言われる「AIは幻想だった」という評価を乗り越えるべく「AIはデジタルだ」の見地から、米国の非営利団体OpenAIが開発した巨大言語モデルGPT-3を例にして、これから見えるイノベーショントレンドを共有する。
Business Environment of Deep Tech AI Startups
There is no naturing environment for deep-tech startups to grow in Japan. I would like to point out the following issues: (1) insufficient development of innovator education system, (2) lack of human resources on the venture capital side, and (3) lack of development of receptors on the large enterprises side. In order to overcome the reputation of "AI was an illusion." from the viewpoint of "AI is digital.", we will share the innovation trends to be seen using the huge language model GPT -3 developed by the non-profit organization OpenAI in the United States as an example.
日本ではディープテック・スタートアップが育つ環境がない。1. イノベーター人材育成が不十分、2.ベンチャーキャピタル側の人材欠如、3. 大企業側レセプターが未発達という課題を指摘した上で、今後どうすれば良いのかという未来志向の議論をしたい。巷間で言われる「AIは幻想だった」という評価を乗り越えるべく「AIはデジタルだ」の見地から、米国の非営利団体OpenAIが開発した巨大言語モデルGPT-3を例にして、これから見えるイノベーショントレンドを共有する。
Business Environment of Deep Tech AI Startups
There is no naturing environment for deep-tech startups to grow in Japan. I would like to point out the following issues: (1) insufficient development of innovator education system, (2) lack of human resources on the venture capital side, and (3) lack of development of receptors on the large enterprises side. In order to overcome the reputation of "AI was an illusion." from the viewpoint of "AI is digital.", we will share the innovation trends to be seen using the huge language model GPT -3 developed by the non-profit organization OpenAI in the United States as an example.
Introduction to behavior based recommendation systemKimikazu Kato
Material presented at Tokyo Web Mining Meetup, March 26, 2016.
The source code is here:
https://github.com/hamukazu/tokyo.webmining.2016-03-26
東京ウェブマイニング(2016年3月27)の発表資料です。すべて英語です。
Open IoT platforms for Smart Services (Japanese)FIWARE
'Open IoT platforms for Smart Services'.
Presentation (in Japanese) by Ulrich Ahle, FIWARE Foundation CEO.
Shown at the OpenStack Days in Tokyo (21/07/2017)
Generative AI: Redefining Creativity and Transforming Corporate LandscapeOsaka University
The advent of Generative AI is redefining the boundaries of creativity and markedly transforming the corporate landscape. One of the pioneering technologies in this domain is the Reinforcement Learning from Human Feedback (RLHF). Combined with advancements in LLM (Language Model) has emerged as a notable player. LLM offers two primary interpretations: firstly, as a machine capable of generating highly plausible texts in response to specific directives, and secondly, as a multi-lingual knowledge repository that responds to diverse inquiries.
The ramifications of these technologies are widespread, with profound impacts on various industries. They are catalyzing digital transformation within enterprises, driving significant advancements in research and development, especially within the realms of drug discovery and healthcare. In countries like Japan, Generative AI is heralded for its potential to bolster creativity. The value generated by such AI-driven innovations is estimated to be several trillion dollars annually. Intriguingly, about 75% of this value, steered by creative AI applications, is predominantly concentrated within customer operations, marketing and sales, software engineering, and R&D. These applications are pivotal in enhancing customer interactions, generating innovative content for marketing campaigns, and even crafting computer code from natural language prompts. The ripple effect of these innovations is palpable in sectors like banking, high-tech, and life sciences.
However, as with every innovation, there are certain setbacks. For instance, the traditional business model of individualized instruction, as seen in the context of professors teaching basic actions, is on the brink of obsolescence.
Looking ahead, the next five years pose pertinent questions about humanity's role amidst this technological evolution. A salient skillset will encompass the adept utilization of generative AI, paired with the discernment to accept or critique AI-generated outputs. Education, as we know it, will be reimagined. The evaluative focus will transition from verifying a student's independent work to gauging their ability to produce content surpassing their AI tools. Generative AI's disruptive nature will compel us to re-evaluate human value, reshaping the paradigms of corporate management and educational methodologies
To be or not to be an academic, big enterprise, startup job that is the qu...Osaka University
"Who said it first is not important." Who gets there first is."
(Takeo Kanade, Circa 1990s)
Before joining a Big Enterprise, Check these
Empathy with the company's vision and mission.
Senior management prepares their own presentation materials (with high IT literacy).
There are executives who joined the company mid-career from outside.
There is a good employee training program.
There are many retired employees who are active after leaving the company.
There is an organization that integrates marketing, development, and operations.
There are no academic cliques.
The most conservative part in a company is mediocre experts who love status quo. Top tier experts tend to climb up mountains from one peak to another peak, so as to explore new ideas and products. You must move also from one to another one. It is said,
“You can raise the bar or you can wait for others to raise it, but it’s getting raised regardless.” Raise your bar higher enough no to succeed now but in the future eventually. Your life is counted hoe many oh-shit moment you experienced. Gotta run.
7. 課題 解決策 価値提案 圧倒的な優位性 顧客
主要指標 チャネル
コスト構造 収⼊入
①②
③
④
⑥
⑦⑧
⑨
• Few algorithm
advances in last 30 yrs
• Top algorithm used for
speech, vision,
language is the
• convolutional neural
network.
• Fundamental
architecture hasn’t
changed since 1980s
• Most performance from
increased CPU and
memory
• Requires millions of
labeled training
examplesNot
generative, can’t do
fine segmentation, etc
受託業務における収
益
ソリューション提供
の精度度、速度度
⼈人件費
7
• Producing precise explanations
require feedback
• connections. (Brain has lots of
them)
• Feed-forward and feedback
work simultaneously
• In contrast, neural networks are
purely feedforward.
Feedback connections
以下の分野の自動化
The world currently pays
humans to do jobs
computers
should do:
Manufacturing - Inspect
and manipulate objects.
Logistics - Drive trucks,
planes, and boats.
Retail - Restock shelves
and ring up merchandise.
Food - Cook and serve
meals.
Agriculture - Plant and
harvest food
• Generative probabilistic graphical
model.
• Currently performing near human-level
on shape recognition.
• On track to reach human-level object
recognition.
• Requires very few training examples.
• Very fast, memory efficient.
• Can be applied to vision, audio,
language, motor, etc
ロボティクス
が利利⽤用できる
接客事業、
建設業
製造業
以下のロードマップで開
発継続.
shape
texture
motion
motor
language concepts
未確定
B2Bチャネル
全世界から選別した30⼈人
ほどのPh.D.
No plans to commercialize current tech, but as a fallback plan
OCR is a $400M/year industry.
現時点での収入はなし.以下の投資のみ.M&A狙いは自明か.
• 1.2M Seed. Founders Fund, Felicis VC, Open Field Capital,
• others. Oct 2010.
• 15M Series A. Dustin Moskovitz, Founders Fund, Open Field
• Capital, others. Jul 2012.
• 40M Series B. Mark Zuckerberg, Elon Musk, Jeff Bezos,
• Jerry Yang, Dustin Moskovitz, Formation 8, Khosla Ventures,
• Founders Fund, Felicis VC, others.
技術志向の強いスタートアップ.脳の視覚機能実現を当⾯面狙う.
Vicariousに対する私の⾒見見⽴立立て
8. Building
Height Oil
Tank
Farms Ag
Areas
and
Yield Tract
Housing
Clouds
&
Haze Development AirplanesWater
衛星写真解析による業態分析
統合化事例例
8
Orbital Insight
23. まともな序⽂文との共通性
論論⽂文の序⽂文の構成要素にまともな「社会への洞洞察」を⼊入れろということ
• 研究の必要性,意義
• 従来のその分野の研究状況
• 研究の⽬目的,扱う範囲
• オリジナリティを主張する範囲
Introduction is a brief, outline of the “Problem” (or Position). It is designed to
gain the prospective reader's interest and his/her understanding, so that he
will be inclined or compelled to follow your line of arguments(present in
body). Tell the reader why your paper is important or show him/her how to
use it! 23