TASK FOR THE COURSE ASSIGNMENT
Course: Deep Learning for Cognitive
Computing: Theory
The major objective:
During
the assignment you are supposed to find, select, learn, apply, and evaluate a
diverse set of available online artificial intelligence (AI), machine learning
(ML), cognitive computing (CC), and Generative AI (GenAI) services to design
the task below, which is Cognitive Profile of your own (possible choice,
which requires careful design with respect to privacy) or of someone else
(publicly known person or celebrity) based on available and open data from the
web.
The task:
The
task for the assignment would be creation and visualization of your own Cognitive
Profile using available online demos of the cognitive computing
services. You are supposed to find yourself or/and use some from the list of
available online services (see below few examples), carefully check their privacy
policies and potential privacy issues (see the note below), try some of their
demos and design your personal profile in a similar (preferably better) way as
the partial profiles of Vagan, Pekka, Donald and others were presented in https://ai.it.jyu.fi/vagan/DL4CC_Part-1.pptx.
(Notice that in this huge file, the
slides, which contain examples related to your assignment, are marked with the
white clickable letter A within violet
circle). For designing your profile, you are supposed to take samples of your
chosen person publicly available (and not containing sensitive information)
writings (academic texts, blogs, etc.), speeches, photos, and videos as inputs
for these services and then embed the outcomes to your Cognitive Profile
represented as the PowerPoint presentation. At the end of the presentation,
provide a few conclusive slides aiming to answer the questions:
·
How powerful are the services (AI/ML/CC/GenAI) used? What are their
strengths and weaknesses? Which of them do I like most and why?
·
What could be potential use-cases
and future applications, which may benefit from such cognitive profiles as have
been designed?
·
What could be potential dangers associated with exposing such profiles
publicly and why?
NOTE:
The quality of the assignment will depend on the diversity and relevance
of the AI/ML/CC/GenAI technologies explored, the quality of the resulting Cognitive
Profile, the creativity of the selected use-cases, and (importantly) the
student’s critical evaluation of the capabilities, limitations, ethical
implications, and privacy risks of the technologies used.
While addressing the conclusive questions above you are
supposed to provide your own thoughts, and the use of the AI/ML/CC/GenAI tools
and services is not allowed.
Please
name the resulting PowerPoint (.pptx) file as Firstname_Familyname.pptx to your
personal Web space (e.g., the one provided by the university on users.jyu.fi)
and send the link to the course instructor (via university email address)
before the deadline. The more different details you include to
the profile, the more different services you will find and properly use for
that, and the more interesting use-cases you invent - the better profile you
will get. In this course (Deep Learning for Cognitive Computing: Theory) you are
not supposed (but you may if you want and capable) to program anything or use
programmable APIs to make your cognitive profile; the use of online demos
through GUIs will be enough.
=====================================================================
=====================================================================
Personal Data Security and Privacy Note
Please read this note carefully before starting the
actual work on the assignment. Privacy and data security take precedence over the
completeness or sophistication of your Cognitive Profile.
This
assignment involves experimenting with online AI/ML/CC/GenAI services provided
by third parties. Such services may process, store, retain, or otherwise use
information that you upload to them, and their privacy practices may differ
significantly. Some services may also operate outside the European
Union/European Economic Area. Therefore, you must think about data
privacy not only when storing and submitting your final assignment, but also when
uploading any input data to an online AI service.
1. You are NOT required to expose your own personal
data
Although
creating a Cognitive Profile of yourself is an inspiring option, you are
explicitly allowed to create a profile of a publicly known person instead.
You are also
encouraged to use, whenever appropriate:
· publicly available material about a public figure;
· non-sensitive examples of your own work;
· synthetic or artificially generated examples;
· anonymized or otherwise appropriately modified material; or
· other material that does not reveal private or sensitive information.
You must never upload personal,
sensitive, confidential, or biometric information to an external AI service
merely to obtain a better grade or a more complete Cognitive Profile.
In particular, you should exercise special
caution with face images, voice recordings, videos, biometric information,
documents containing personal identifiers, private correspondence, health information,
financial information, passwords, authentication information, or other
sensitive data.
If a
particular AI service requires uploading data that you are not comfortable
providing, do not use that service. Find another service providing a
similar capability, use a different type of input, or use a public/synthetic
subject instead. There is no requirement to use every service listed in the
assignment.
2. Exploring many AI services does NOT mean uploading
personal data to many services
The assignment
encourages you to explore a diverse selection of AI/ML/CC/GenAI services
because comparison of different technologies is an important part of the
learning objective.
However, the
instruction to use “as much as possible” or “as many as possible” services must
not be interpreted as an instruction to distribute your personal data among as
many online providers as possible.
The quality of
the assignment depends primarily on the diversity, relevance, critical
evaluation, and creativity of the technologies and cognitive features you
investigate, not on the number of external services to which you submit
personal data.
You should
therefore select services responsibly and consider their privacy practices
before uploading any material.
3. Consider the privacy policy of an external AI
service before using it
Before
uploading personal material to an online AI service, consider at least:
·
what data the service
collects;
·
whether uploaded files are
retained and for how long;
·
whether the data may be used
for training or improving AI models;
·
whether data may be shared
with third parties;
·
whether data may be
transferred outside the EU/EEA;
·
whether the provider offers
meaningful deletion or other data-subject rights;
·
whether the service requires
an account and what information is required for registration; and
·
whether you are comfortable
accepting these conditions.
Do not assume that a service is safe
simply because it is listed as an example in this assignment. The list of services is provided for educational exploration and may
contain outdated services, changed policies, or services whose privacy
practices are not appropriate for your particular data.
Likewise, the
presence of a service on the list does not constitute an endorsement by the
course or the university of that service’s privacy policy or GDPR compliance.
If the privacy
conditions of a service are unclear, unusually broad, or otherwise unacceptable
to you, simply choose another service.
For additional
background, you may consult the European Data Protection Board’s Opinion
28/2024 on certain data protection aspects related to the processing of
personal data in the context of AI models, which discusses, among other
things, personal data, anonymity, legal bases for processing, and the
implications of using personal data in AI systems:
EDPB
Opinion 28/2024 on AI models
A shorter
explanation of the same opinion is also available from the EDPB:
EDPB:
AI models and GDPR principles
4. Privacy of the final assignment and privacy of data
submitted to AI services are two different issues
Keeping your
final PowerPoint file private does not automatically protect information
that you previously uploaded to third-party AI services.
Therefore, you
should apply privacy protection at both stages:
A.
During experimentation with external AI services:
Do not upload
private, confidential, sensitive, or unnecessary personal information. Use
public, synthetic, anonymized, or otherwise appropriately selected material
whenever possible.
B.
When preparing and submitting the final assignment:
Make sure that
the PowerPoint itself does not contain private or sensitive information that you
do not want to disclose to the teacher or that should not be included in the
assignment.
5. Storage and delivery of the assignment
Both students
and teachers must ensure that the assignment content and the student-to-teacher
storage and delivery procedure comply with applicable university policies and
data-protection requirements. Please carefully read the university’s current
policies and guidance concerning information security and personal data
protection.
The preferred
delivery method is:
1.
Email the assignment from your university email
account to the teacher’s university email account, preferably as an attachment. If the file is larger than 10 MB, use an
appropriate secure university-provided cloud/storage service for transferring
it.
2.
Alternatively, upload the
file to a password-protected subdirectory of your personal university web space,
if such a facility is available and permitted by university policy, and provide
the teacher with the link and a one-time password for downloading the file.
Do not place
the assignment in a publicly accessible directory.
After the
teacher confirms successful download, you may remove the file from the cloud
service or web directory.
The teacher
will not intentionally disclose your assignment to other persons. The submitted
assignment will be stored securely for approximately two years for
course/academic administration purposes and will be deleted after your graduation,
subject to applicable university retention requirements.
6. An important principle for this assignment
Responsible experimentation is part of
the assignment.
Discovering
that an AI service can infer surprisingly detailed information about a person
is itself an interesting result. Discovering that obtaining such an inference
may require surrendering data that you would not want a third party to possess
is also an important result.
Therefore, you
are expected to exercise your own judgement about what data you are willing to
provide to each service. You will not be penalized for refusing to use a
service because of legitimate privacy or data-security concerns.
A smaller
Cognitive Profile based on responsibly selected data is preferable to a more
extensive profile created by unnecessarily exposing personal or sensitive
information.
Finally,
please remember that publicly available information is not automatically
risk-free information. The fact that a photograph, speech, text, or video
can be found publicly does not necessarily mean that it should be uploaded to
every AI service. Consider both the nature of the data and the purpose and
conditions of the service before using it.
=====================================================================
=====================================================================
SOME EXAMPLES OF Cognitive Computing Services Online (Notice, be careful
and patient, the list below could be incomplete and include outdated services
or services with potential privacy issues):
TEXT-TO-TEXT
A. Summarization
http://textsummarization.net/text-summarizer ;
https://deepai.org/machine-learning-model/summarization
;
https://quillbot.com/summarize ;
https://algorithmia.com/algorithms/nlp/Summarizer
;
https://chat.openai.com/chat
;
B. Generation
https://talktotransformer.com/ ;
https://deepai.org/machine-learning-model/text-generator
;
https://www.poem-generator.org.uk/
;
https://app.inferkit.com/demo
;
https://chat.openai.com/chat
;
C. Translation
https://translate.google.com ;
https://www.deepl.com/translator ;
https://www.ibm.com/demos/live/watson-language-translator/self-service/home
;
https://chat.openai.com/chat ;
D. Question & Answer
https://www.pragnakalp.com/demos/BERT-NLP-QnA-Demo/ ;
https://deeplearninganalytics.org/demos/
;
https://dida.do/demos/question-answering
;
https://chat.openai.com/chat ;
TEXT-TO-SPEECH
https://www.ibm.com/demos/live/tts-demo/self-service/home
;
https://azure.microsoft.com/en-us/products/cognitive-services/text-to-speech/
;
https://cloud.google.com/text-to-speech#section-2
;
https://www.naturalreaders.com/online/
;
https://www.descript.com/ai-voices
;
TEXT-TO-SEMANTICS (understanding, tone, emotions, personality, essential content, etc.)
https://www.ibm.com/demos/live/natural-language-understanding/self-service/home
;
https://aidemos.microsoft.com/text-analytics
;
https://cloud.google.com/document-ai
;
https://www.intelligenceapi.com/demo/
;
https://deepai.org/machine-learning-model/sentiment-analysis
;
https://deepai.org/machine-learning-model/text-tagging
;
https://www.entitysearch.io/#demo ;
https://ibm-pi-demo.mybluemix.net/
;
http://www.blablameter.com/index.php
;
SPEECH-TO-TEXT
https://www.ibm.com/demos/live/speech-to-text/self-service/home
;
https://cloud.google.com/speech-to-text
;
https://cloud.google.com/speech-to-text/?hl=en ;
SPEECH-TO-SPEECH
SPEECH-TO-SEMANTICS
https://aidemos.microsoft.com/luis/demo
;
IMAGE-TO-SEMANTICS
https://cloud.google.com/vision/ ;
https://www.clarifai.com/models/image-recognition-ai
(more: https://www.clarifai.com/computer-vision
);
https://face-api.sightcorp.com/demo-basic/
;
https://vision-explorer.allenai.org/detection
;
https://developer.opencv.fr/ui/#/onboard/demo
;
TEXT-TO-IMAGE
https://deepai.org/machine-learning-model/text2img
;
https://hotpot.ai/art-generator ;
https://creator.nightcafe.studio/ ;
https://catalog.ngc.nvidia.com/orgs/nvidia/teams/playground/models/sdxl ;
https://deepai.org/machine-learning-model/text2img ;
IMAGE-TO-IMAGE
https://deepai.org/machine-learning-model/image-editor ;
https://www.artguru.ai/swap-face/# ;
https://demo.changemyface.com/ ;
https://age-and-lifestyle.changemyface.com/ ;
https://ailab.wondershare.com/tools/aging-filter.html
;
https://www.fotor.com/images/create ;
https://deepai.org/machine-learning-model/image-similarity
;
https://betaface.com/demo.html ;
https://reface.ai/unboring/face-swap ;
https://www.photoroom.com/tools/remove-object-from-photo ;
https://blackandwhite.imageonline.co/ ;
https://playback.fm/colorize-photo ;
https://palette.fm/color/filters ;
https://vanceai.com/old-photo-restoration ;
https://www.nvidia.com/research/inpainting/ ;
https://3dthis.com/3dface.htm ;
https://openai.com/blog/glow/ ;
https://deepai.org/machine-learning-model/colorizer
;
http://demos.algorithmia.com/colorize-photos/ ;
https://deepai.org/machine-learning-model/toonify
;
https://www.befunky.com/create/photo-to-cartoon/
;
https://deepai.org/machine-learning-model/torch-srgan
;
https://deepai.org/machine-learning-model/waifu2x
;
https://deepai.org/machine-learning-model/deepdream
;
https://deepdreamgenerator.com/ ;
https://deepai.org/machine-learning-model/fast-style-transfer
;
https://deepai.org/machine-learning-model/neural-style
;
http://cs.stanford.edu/people/karpathy/convnetjs/demo/image_regression.html
;
https://thispersondoesnotexist.com/ ;
https://looka.com/logo-maker/ ;
https://ai.cloudinary.com/demos/fill
IMAGE-TO-VIDEO
https://app.tokkingheads.com/homepage
;
https://www.myheritage.com/deep-nostalgia/ ;
https://sketch.metademolab.com/canvas
SPEECH-TO-VIDEO
https://seamless.metademolab.com/expressive/
IMAGE-TO-TEXT
https://milhidaka.github.io/chainer-image-caption/
;
https://jah.outsystemscloud.com/ImageToTextDEMO/
;
TEXT-TO-VIDEO
https://studio.d-id.com/?utm_content=photo2video
;
VIDEO-TO-TEXT
https://www.veed.io/tools/video-to-text
;
https://www.audiotype.org/en/transcribe/video/ ;
https://vi.microsoft.com/en-us ;
https://www.readtheirlips.com/
VIDEO-TO-AUDIO
https://audio-extractor.net/ .
VIDEO-TO-VIDEO
https://sam2.metademolab.com/demo
AUDIO-TO-AUDIO
https://audiobox.metademolab.com/storymaker/demo
ChatGPT and other chatbots and multiservice providers
https://chatgptonline.net/ [tutorial: https://metaroids.com/learn/what-is-chatgpt-beginners-guide-to-using-the-ai-chatbot/]
https://copilot.microsoft.com/
https://bing.com/
(chat)
https://portal.vision.cognitive.azure.com/gallery/featured
https://ai-service-demos.go-aws.com/
[useful tutorials: https://www.simplilearn.com/tutorials/chatgpt-tutorial/chatgpt-alternatives
]
and many-many other services/demos
online and their categories
-------------------------------------------
IMPORTANT
NOTICE!
Taking into account that
companies, who are providing these demos (listed above), often update their services
or change links to them, you are expected to use Google to find these (with
broken links) or similar services/demos online. It will be appreciated if you
will find interesting new demos for your assignments (especially, new
categories of revenant demos) online, which are not listed above. Please
respect the Privacy Note above while choosing the services or finding the new
ones.
-------------------------------------------
Read more about AI-as-a-Service providers here:
https://encord.com/blog/ultimate-guide-ai-as-a-service/
https://redresscompliance.com/exploring-cloud-based-machine-learning-platforms/
Additional general and useful references:
- Online textbooks containing theories behind the cognitive computing
services:
Goodfellow, I., Bengio, Y., &
Courville, A. (2016). Deep Learning, MIT Press, 787 pp. (http://www.deeplearningbook.org)
Michael Nielsen (2017). Neural Networks and Deep
Learning.
(http://neuralnetworksanddeeplearning.com/)
Petersen, P., & Zech, J. (2024). Mathematical Theory of Deep
Learning. arXiv preprint
arXiv:2407.18384. (https://arxiv.org/pdf/2407.18384
)
Sumithra, M. G., Dhanaraj, R. K., Iwendi, C., & Manoharan, A. M.
(Eds.). (2022). Deep Learning for Cognitive Computing Systems:
Technological Advancements and Applications (Vol. 7). Walter de
Gruyter GmbH & Co KG. (https://www.scribd.com/document/859454267/Deep-Learning-For-Cognitive-Computing-Systems-Technological-Advancements-And-Applications-Mg-Sumithra-pdf-download
)
https://www.ayadata.ai/top-10-books-on-machine-learning-and-ai/
https://digitaldefynd.com/IQ/top-books-for-learning-artificial-intelligence/