• No results found

Tillämpad AI. Staffan Truvé

N/A
N/A
Protected

Academic year: 2022

Share "Tillämpad AI. Staffan Truvé"

Copied!
21
0
0

Loading.... (view fulltext now)

Full text

(1)

Tillämpad AI

Staffan Truvé

• Varför är jag här?

• CTO och medgrundare, Recorded Future – ”AI for Intelligence”

• Ordförande, AI Sweden

• Ingenjörsvetenskapsakademien (IVA) – ”Peak Human”

(2)

Accelerating applied AI

in Sweden

(3)

Accelerating the use of AI for the benefit of our society, our

competitiveness, and for

everyone living in Sweden

(4)

National center for applied AI

Neutral,

non-profit and broadly

funded

Here to help AI help you

Accelerating applied AI Swedish industry and society

Backed by Swedish government, by public and private sector

Strengthen and

transform operations through the use of AI

(5)
(6)

Invest together – Share with all

Cross-sector

(7)

Examples of what we do

Value creat ion

AI expert ise and t echnology Dat a and

inf rast ruct ure

Business underst anding

Executive AI Accelerator Startup program

Legal Expert Group Data Factory solutions

- Compute and storage capacity - Infrastructure testbed

AI Change Agent Program AI for Leaders

Decentralized AI

Language technologies

(8)

Strategic areas of applied AI

Decentralized AI Privacy preserving solutions

Language technologies for Swedish & Other

languages

Multi-modal insights &

integration

Complex systems

management Organization for AI impact

(9)

Edge Lab by partners Expertise & knowledge

sharing Investments by

partners

Infrastructure Manpower

Use case by partners

Automotive

Healthcare

Recycling

Coming use case: Pharma, security, etc

Research & Innovation partners to be invited

International collaboration – Early discussion initiated

Example: Decentralized AI

(10)

AI for Intelligence

Staffan Truvé PhD

CTO & Co-Founder Recorded Future

(11)

Där all världens information,

transaktioner, IP och pengar bor och flyttar på sig

Där cyberattacker planeras, utförs och kapitaliseras

GOTT ONT

(12)

Bygg en riskorienterad tvilling av världen - i Internetskala och i realtid

Organisera för analys -

av människor och algoritmer

Möjliggör “riskanalys-kentaurer”

AI för automatisering och förstärkning

Grundläggande idéer

(13)

Structured data from technical sources Unstructured text sources

Hur bygger man en digital tvilling?

Source: https://krebsonsecurity.com/2020/10/fbi-dhs-hhs-warn-of-imminent-credible-ransomware-threat-against-u-s-hospitals

Security Intelligence

Graph

(14)

Open Source Collection

Risk Scoring

Targeted Collection Victim Identification

Alerting

(15)

AI @ Recorded Future

Kunskapsrepresentation – Security Intelligence Graph

Textanalys / NLP

Risk Scoring / Anomalidetektering

Prediktiv Analys

(16)

Infra- structure

Victims

Capabilities Adversaries

Red Delta

167.88.

180.5 PlugX

codegreen.ae

185.239.

226.65

167.88.

180.132 crcl.lu

6aad032a0 84de893b0 e8184c17f0

376a

CVE 2012-

0158 CVE

2017- 0199 APT

10

Axiom Hacking Group

whitewall.

top

46.*.*.100

Vatican City

Security

Intelligence

Graph

(17)

Collection

Text Analytics/

NLP

Data Analytics

RULE BASED ML BASED

Vem? Vad? Var? När? Varför?

Text classification

Language classification

Topic classification

Entity recognition

Event classification

Role assignment

Temporal analysis

Sentiment analysis

(18)

Country: Russia Company: Nuance Industry: Hospitals Malware: Petya

Country: United Kingdom

Event Extraction

(19)

Source: https://xkcd.com/195/

Exempel: prediktering av IP-risk

(20)

Slutsatser

• AI kan verkligen ge företag unika konkurrensfördelar

• Inte en lösning i sig, utan som del i en systemlösning

• AI erbjuder en portfölj av verktyg – representation, logik, ML

• Kentaurer: människa + maskin

• Vi har knappt ens börjat…

(21)

Staffan Truvé

truve@recordedfuture.com

References

Related documents

It uses the solutions and algorithms developed within Natural Language Generation (NLG) applications to efficiently process data and adapt the presentation of text content to a

This thesis explores the optimal ways in which natural language generation techniques can be brought to bear upon the problem of processing a structured body of information in

This thesis aims to identify the optimal ways in which natural language generation techniques can be brought to bear upon the problem of processing a structured body of information

In this paper we compare and assess four freely available cross-sectional time-series data sets in terms of their information on the ballot structure, district structure and

Analysis of total organic carbon (TOC) and the water content for the allochthonous sediment (Chapéau d’Uvas) and the autochthonous sources: phytoplankton and the different species

Semi-structured interviews have been used in this thesis. A semi-structured interview is not as strict compared to a structured interview. This means that the interviews are

We can suppose that the FRCE target parameters will be better than with the current aFRR prescription which to a mean scheduled aFRR level about 50 MW smaller than with the Δ 30

Following a discussion of the conceptualization and measurement of legitimacy beliefs per se, the chapter has assessed five types of individual-level explanations: in terms of