FRSC-3100: Exam Review

Disruptive Technology

An innovation that significantly alters how consumers, industries, or businesses operate.

AI is based on:

Algorithms that mimic human neural networks.

AI on its most basic level can:

Identify patterns, predict sequences and/or generate data in a fashion that can appear life-like in its speed and sophistication.

Generative AI

Focuses on understanding patterns and structure in data and using that to create new data that looks like it.

Examples of Generative AI

• Writing blocks of text

• Writing lines of code

• Creating photo-realistic images

Predictive AI

Focuses mainly on classification, learning the difference between “things”.

Examples of Predictive AI

• Recommendation engines used by Netflix or Amazon to distinguish between things you might want to watch/buy and things you’re unlikely to be interested in.

• Navigation apps to distinguish between good routes from A to B and one you should avoid.

Artificial Narrow Intelligence (ANI)

Specialized to the function in which it has been developed. 

Artificial General Intelligence (AGI)

Describes AI as smart as humans, capable of planning, reasoning, and problem-solving as fast as,or perhaps faster than the human brain.

Artificial Super Intelligence (ASI)

The point at which computers possess intellectual capacity far greater than that of human beings with the capacity for social skills and general knowledge that would increase exponentially over time.

SPOT terrestrial robot

• SPOT CAM captures spherical images and comes with 30X optical zoom for detailed inspections.

• SPOT CORE provides dedicated processing for applications requiring on-robot computation.

• SPOT ARM enables mobile manipulation for tasks like opening doors and grasping objects.

• SPOT GXP provides regulated power and an Ethernet port

Multi-Modal Mobility Morphobot (M4)

• Terrestrial mobility on wheels

• Walking, crouching, wheel rolling, and tumbling capabilities

• UAS flying capability

Neural Radiance Field (NeRF)

Computer-generated 3D models mimicking lighting dynamics for near photo-realistic representations.

ClimateNeRF

Physically-based neural rendering for extreme climate synthesis.

SLAM

A handheld device that scans and maps the dimensions and layout of a scene in real time via a rotating laser.

Photogrammetry

Compilation of photos and videos to map and visualize a scene.

Machine Learning

Utilizes data and algorithms to mimic human learning, used in tasks like medical diagnosis and facial recognition.

 The Campbell Report

• 1996 regarding R v Paul Bernardo

• Issues: Gaps in the investigation/CFS processes

• Result: Major Case Management

The Kaufman Report

• 1998 regarding R v G.P. Morin

• Issues: gaps in the investigation/CFS bias, junk science

• Result: Call for forensic scientists to be objective, independent, and accurate

The Goudge Report

• 2008 regarding Dr. Charles Smith

• Issue: Faulty forensic pediatric pathology with erroneous conclusions

• Result: Commitment to training, collaboration and peer review

The Hart House Report

• 2013 Forensic Science in Canada: A Report of Multidisciplinary Discussion

• Issue: A patchwork of contributions

• Result: Call for improved higher training, guidelines, standards, certification, accreditation and ethics covering nine forensic disciplines

  • Pathology, Anthropology, Odontology, Nursing, Entomology, Physical sciences, Toxicology, Biology, and PsychiatryLarge Language Models

Large Language Models

AI program designed to understand, generate, and work with human language on a large scale.

Natural language Processing (NLP)

Involves enabling computers to understand, interpret, and respond to human language in a way that is both meaningful and useful.

Algorithm

A simple set of rules.

Deep Learning

Subset of machine learning using neural networks for analyzing unstructured data like images and text.

AI Training Data Issues:

• AI has no semantic understanding of words

• AI is only as smart as the data it is trained on and the feedback it receives

• Biased data in language and photos

Feature-Based Facial Recognition

Measures the distance between facial features and creates geometric relationships.

Applications of Feature-Based Facial Recognition:

• Often used in systems where high precision is required and in situations where the facial expressions  might change

• Automated border control systems, passports and smartphones

Appearance-Based Facial Recognition

Use holistic information from the face and analyze the entire facial image through Principal Component Analysis or Linear Discriminant Analysis.

Applications of Appearance-Based Facial Recognition:

• Commonly used in environments where facial conditions are more controlled

• Passport or ID verification systems and social media photo tagging.

Knowledge-based Facial Recognition

Designed to identify suspicious or abnormal behaviours in real-time based on predefined rules or knowledge about facial expressions and behaviours.

Pros of Facial Recognition:

• Enhanced security and safety

• Efficiency in policing

• Convenience in consumer applications

• Support in forensic work

Cons of Facial Recognition:

• Privacy concerns

• Potential for abuse

• Bias and inaccuracy

• Legal and regulatory challenges

• Effects on civil liberties

Automated Fingerprint Identification System (AFIS)

Biometric system storing fingerprint and palm print images for identification purposes.

AFIS System

Automated Fingerprint Identification System that uses advanced algorithms to search fingerprint and palm print records in databases.

Each AFIS record contains:

• 10 rolled images

• 10 plain images

• 6 palm images

Each AFIS record is stored in two files:

• 1 image file

• 1 minutiae map

How does AFIS interpret data?

• It identifies features like ridge endings, bifurcating ridges, and large ridge dots.

• These features are located and noted with their X-Y coordinates and theta vectors and are reduced to mathematic values.

• AI technology measures the physical distances between features.

Third Level Features

Additional features beyond minutiae used in modern AFIS systems for more accurate database searches.

The problems AI tries to solve in AFIS:

• There are subtle differences in each friction ridge record, including those from the same person.

• The AI uses advanced mathematics to anticipate these differences while at the same time searching the constellation of features within various large databases.

Limitations of AI in AFIS:

• Ethics: AI does not function ethically just yet

• Privacy: AFIS records must be scrutinized by people for legality of use

Deviation/Distortion

Changes in fingerprint ridge details due to injuries or deliberate distortions that AFIS must accommodate.

What is CPAP?

• 3D crime scene documentation procedure

• Capture, Process, Analyze, Present

UAVs in CSI

Offers aerial perspectives that were once inaccessible, enabling rapid data collection and comprehensive scene analysis.

Laser Scanning in CSI

Delivers precise point cloud data, facilitating the creation of detailed 3D models crucial for forensic examinations.

NeRF in CSI

Driven by deep learning algorithms and synthesizes immersive 3D reconstructions from 2D images, providing unparalleled fidelity in crime scene visualization.

Videogrammetry in CSI

Extracts 3D information from video footage, aiding in event reconstruction and spatial analysis.

Human Computer Interaction (HCI)

The discipline is concerned with the design, evaluation, and implementation of interactive computing systems for human use and the study of the major phenomena surrounding them.

Transhumanism

The belief that the human race can evolve beyond the current physical and mental limitations, especially by means of science and technology.

Man resurrects childhood imaginary friend (microwave oven) using AI.

Then it tried to kill him.

Salt

The first fully AI generated movie

Google Research Lumiere

A Space-Time Diffusion Model for Video Generation that can turn text prompts into a video and image prompts into a video.

Evidentiary issues in AI:

• Fake text

• Fake images

• Fake people

• Fake videos

Why was Guy Paul Morin arrested for the murder of Christine Jessop?

• Neighbour, lived close to the Jessop’s

• Strange fella, socially inept and quiet

• Didn’t help search for Christine

• Didn’t attend the funeral

What genealogy lab was selected for the Jessop case?

The Othram INC lab in Texas.

Where in Canada were the two distant relative Gedmatch results found in the Jessop case?

Belleville, Ontario.

What genealogy sites allow law enforcement use?

Gedmatch and FTDNA.

What are some open data sources for genealogy?

• Death notices

• Instagram

• Facebook

• Newspapers

Centimorgan

Unit measuring genetic linkage used in genealogy to determine relationships based on DNA matches.

Who was Christine Jessop’s killer?

Calvin Hoover.

The autosomal STR DNA results are estimated to be greater than __________ times more likely to originate from Calvin Hoover than an unknown person unrelated to him.

One trillion.

What are the two main DNA databases?

• The National DNA Databank (NDDB)

• The Local DNA Database (CFS)

Challenges with DNA evidence:

• Issues regarding secondary and tertiary transfer of DNA

• Increased issues with collection (cross contamination)

Examples of DNA contamination sources explored in scholary articles:

• Cut resistant leather slash gloves worn by officers

• Multi-use fingerprint brushes in forensic casework

Fingerprint Evidence

Involves latent fingerprint development, collection of known impressions, identification of deceased, search through AFIS, comparison, and presentation of findings.

Fingerprint development techniques are chosen for:

• Surface: Porous vs non-porous

• Fingerprint matrix: Blood, sweat, dust

• Contrast and photography: powder, fluorescent powder, cyanoacrylate fuming

Vacuum Metal Deposition (VMD)

Depositing very thin layers of metal on to a surface in a vacuum chamber to visualize latent fingerprints or touch marks. Zinc, gold, and silver are commonly used.

Blind Verification

Process of verifying fingerprint comparisons without knowledge of the initial examiner's conclusions to ensure accuracy.

Fingerprint Development Techniques

Methods like powder, fluorescent powder, and cyanoacrylate fuming used based on surface type and fingerprint matrix.

ALS Examination

Involves the use of lasers and forensic light sources for examining fingerprints on various surfaces.

Technology-based bloodstain pattern analysis

• Eye tracking study of bloodstain pattern analysts during pattern classification

• Automatic classification of bloodstains with deep learning methods

• Automated reconstructions of cast-off patterns based on Euclidean geometry and statistical likelihood

• Blood deposition and drying time estimation

Basics of Virtual Reality

Involves computer-generated immersive environments experienced through sensory stimuli affecting actions within the environment.

Fidelity in VR

Refers to the realism of visuals, accuracy of tasks, and cognitive processes in the immersive environment.

Perceptual Fidelity

Virtual interactions that closely mimic the physical world activate the same neural pathways in the brain.

Feedback Fidelity

In VR, learners make decisions just as they would in the real world, and depending on the level of engagement, these decisions can have direct and immediate positive or negative impact.

Emotional Fidelity

In VR, can invoke a sense of presence that creates real emotional and empathetic responses. These sensory stimulations trigger the brain in much the same way it reacts to actual situations to release endorphins, serotonin and/or dopamine.

Agency in VR

The degree of choice, interaction, ownership, and/or control.

Formula for Presence in VR

Agency + Immersive Environment = Presence

Suspension of Disbelief (SoD)

The first steps in immersion and the formation of Presence involve Suspension of Disbelief.

The Narrative in VR

Immersion is significantly enhanced by a compelling and comprehensive narrative. It contextualizes and helps make sense of the activity and tasks being encountered.

3 DoF

VR headsets with 3DoF only have rotational control. You can look around but not interact with the environment or dictate where to go.

6DoF

VR headsets with 6DoF have both rotational and translation control. Subjects have control over interacting with the scenario and can use haptic technology to enhance the experience.

VR for Training

Used for recruiting, on-boarding, skill development, and testing, except for individuals susceptible to VR sickness.

What are DICE scenarios?

• Dangerous

• Impossible

• Counter-productive

• Expensive

Pros of AI for training:

• Flexibility in accessing training when convenient

• Provides the ability to explore in private

• Asynchronous or synchronous as needed

• Team setting when mutually convenient

• Can be used as a preparatory stage working towards a final hands-on/real-world exam

• Provides a way of experiencing a wider variety of subject situations or scenarios

Cons of AI for training:

• Delayed feedback from instructor

• Delayed tech help

Why VR training?

Improved comprehension, internalization, retention and confidence in students.