Suomeksi - In English
Small notes along the journey. The Finnish and English versions of the notebook won't, unfortunately, keep the same pace.
An interview with Iyad Rahwan about a new discipline he calls "machine behaviour". The discipline is about diving into algorithms and their "behaviour", but without any assumptions on any specific level of intelligence, skipping the consciousness discussion etc.
"Joanna Bryson from the University of Bath, she’s always outspoken against giving machines agency --
But for me, behavior doesn’t mean that it has agency [in the sense of free will]. --"
Facebook launches new tools for better privacy management on the platform. However, immediately critical voices are following:
Guardian: Facebook launches 'clear history' tool – but it won't delete anything
I haven't got time to dive into this, but I try to follow the situation. Will update this when there is more information / I get to test the privacy tool.
// It would be hard to see Facebook making this clear changes and risking their business model. Linked straightforwardly to the main theme "AI Ethics: Data centralization and privacy"
University of Turku has founded an AI Academy (Tekoälyakatemia) to assemble and coordinate education and research related to AI technologies.
One significant novelty is a multidisciplinary minor programme of AI. In addition to computer science, the programme offers courses of philosophy, ethics, social sciences, law and applied medicine.
However, unfortunately the programme is available only in Finnish at the moment.
You can find more information about the AI Academy and AI 25 Programmer from: https://sites.utu.fi/tekoalyakatemia/koulutus/tekoaly25/
This is the first time I read Peter Singer's comments on AI.
He is commenting that possibly sentient/conscious machines should be treated as good as humans and animals, which seems quite logical to his thinking. However, by this simple interview, it feels that he is not really familiar with the topics related to AGI, machine consciousness etc.
In addition, he wisely comments also:
"It is also going to change the social landscape regardless of these machines are conscious. It is certainly going to alter employment. We have to think about what we are going to do when we get replaced by AI. We have to think about how we are going to support that society. Are we going to have a universal basic income? These are very important questions that we will be grappling with over the next decades."
The interview is by Cherwell.org:
https://cherwell.org/2019/05/17/interview-peter-singer/
Still under debate, and often challenged by the computer scientists and engineers, but some definition of AI makes the societal and ethical discussion about AI more valuable.
The definition by the European High-Level Expert Group is a good starting point:
“Artificial intelligence (AI) refers to systems that display intelligent behaviour by analysing their environment and taking actions – with some degree of autonomy – to achieve specific goals.
AI-based systems can be purely software-based, acting in the virtual world (e.g. voice assistants, image analysis software, search engines, speech and face recognition systems) or AI can be embedded in hardware devices (e.g. advanced robots, autonomous cars, drones or Internet of Things applications).”
Autonomous vehicles haven't been the exact focus for my research, but I am following the discussion with a great interest. Few colleagues in the University of Turku are doing research related to the theme, f.ex AALaw project at the Department of Law. In addition, I am taking myself part in the research alliance called RAAS, you can find more about RAAS at autonomous.fi.
An interesting resource for the autonomous vehicles, is the report "Automated and Connected Driving" by the German Ethics Commission of the Ministry of Transport and Digital Infrastructure.
It surprises me how many people are so eagerly and purely interested in making humans (any of genders) obsolete.
Is there a reasoning behind (f.ex the ultimate lazy satisfaction of needs) or is it only the fresh and novel idea?
I am not critizing O'Neil's mid-level point (the gender-centered discussion about sex robots), but the background point "it's OK to think making people or humans obsolete".
https://www.bloomberg.com/opinion/articles/2018-01-04/maybe-sex-robots-w...
I wrote to the blog of Tekoälyaika project (by the Ministry of Economic Affairs):
Words into action: Who is responsible for ensuring the ethics of artificial intelligence?
As in the year 2018 the focus was the creation of AI ethics guidelines, this year should be the year of practical implication and execution of AI ethics.
I warmly recommend John Danaher's podcast http://philosophicaldisquisitions.blogspot.com/p/podcast.html
It offers a great opportunity to hear the opinions of various AI experts, and goes through most of the specific questions related to AI ethics. Danaher's philosophical work is worth of reading, too.
There seems to be various solutions to make people trust autonomous vehicles more. The solutions are quite clever, for example adding moving eyes to the vehicles, to signal that the vehicle has noticed passing pedestrian.
Good examples here: https://www.fastcompany.com/90231563/people-dont-trust-autonomous-vehicles-so-jaguar-is-adding-googly-eyes
Still working with my knowledge of physics, but this lecture by Carlo Rovelli contained interesting thoughts on the nature of time.
"-- and the closest to us, I think, the strong emotional connection with time, the emotion of time, is what time is for us."
https://www.youtube.com/watch?v=-6rWqJhDv7M
It would be interesting to hear what Valtteri Arstila from University of Turku would say about the topic.
https://www.utu.fi/fi/yksikot/soc/yksikot/filosofia/oppiaine/henkilokunta/Sivut/Arstila.aspx
A big recommendation for this review on the earlier and latest AGI safety research. 15 pages just dynamite, 15 pages sources. Read that all and you're really back in the game!
I mostly agree with Professor Yuval Harari. Same old things in the beginning, but really well formulated arguments in the end.
I really don't understand why we are that into making our society more efficient (95%) than thinking these ethical questions (5%).
https://bdtechtalks.com/2018/01/31/yuval-harari-wef-ai-big-data-digital-...
There is a thread in Twitter about good examples of ethical codes in technological companies. An interesting example is from Sage UK, you can find the code here: The Ethics of Code: Developing AI for Business with Five Core Principles
You can answer straight to Alan Winfield via Twitter:
As part of a survey of ethical governance in tech companies I want to collect examples of companies with strong and transparent (i.e. published online) ethical codes + process. Please RT and/or reply with examples of good practice. @robmccargow @KayFButterfield @johnchavens
— Alan Winfield (@alan_winfield) January 19, 2018
Roger McNamee's writing is a bit radical, but comprehensive presentation on what's wrong with the monopoly position of social and internet platforms companies.
- social media platforms' role in political elections
- built-in discrimination in advertising
- exploitation possibilities of these platforms by bad actors
- data centralisation
- regulation of social media and internet platforms (free products business model)
https://washingtonmonthly.com/magazine/january-february-march-2018/how-t...
MIT launches an artificial general intelligence course for this spring. Absolutely worth of following.
There is good proposals for under-appreciated AI achievements and progress in 2017 under the Andrew Ng's tweet:
https://twitter.com/AndrewYNg/status/946487653993562112
A positive techno-economical view on AI is starting to dominate the discussion on AI and its future. That was predictable while it is still the existing paradigm on how we see our world and from the capitalist perspective, there is a lot of money to be made ("Gartner predicts that by 2021, AI augmentation will create $2.9 trillion of business value").
Gartner also put its effort on this lobbying and stating powerfully that already by 2020, AI would be creating more jobs than it eliminates.
I took a quick glance on the material and I didn't find a clear argument on that phenomenon. The strongest reference was " the AI augmentation", which means combining human and machine capabilities. I didn't understand how this would differ noticeably from the current work life or employment problems we face via the automation development
David Graeber's idea on the value of work and bullshit jobs fascinates me a lot. I can see this happening and I hear a lot about this. People aren't doing that much at their jobs. The value-creation has built on something else.
Together with Slatestar Codex's cost disease theory, I feel familiar with this thought. I can see it happening - a lot.
https://www.youtube.com/watch?v=tpoJIkqEXYo
https://www.youtube.com/watch?v=kehnIQ41y2o
Four reasons:
- Science fiction
- Uncanny valley
- Click-bait media
- Fear of the unknown
1. Societal cost: fear of AI slows down the adoption of AI solutions
2. Scientific cost: without research funding, no world-class science
3. Industrial cost: no world-class science, no innovations
How we can change this:
- experts should reach out more to the public about their work
- journalists should dump their click-bait headlines
- we should all be more critical about what we read and see
- most importantly: we need AI education on all levels, free for all
Exponential is not enough: exponential progress (speed) is trumped by exponential increase in problem hardness
- Think of the self-improving system: yourself
Narrow AI: even though computers beat humans in certain tasks (arithmetics, chess, Go...) they are very stupid in many other ways
- There is no ”brain-in-a-jar” AI that learns new skills: it's always a different system
- University education: already strong and growing
- Professional training
- Schools
- Open education for all
Part I: Elements of AI
Starting in May 2018
Three weeks, 2 credit units
Philosophy and history of AI
Basic concepts: Search and planning, games, machine learning, neural networks, signal processing, robotics
”AI literacy”
No programming or maths skills required
Free of charge, advised by Risto Siilasmaa
The chairman of Nokia offers a great, simple and understandable presentation on machine learning. Focusing on the fundamentals and basic procedures of machine learning. The presentation is a really good starting point for anyone not understanding the machine learning method yet.
https://blog.networks.nokia.com/innovation/2017/11/09/study-ai-machine-l...
You can find his presentation from the link below. As most of computer scientist, he don't say anything about ethical and existential risks, but override those as non-scientific arguments. This is probably because the philosophical premises of the computer science.
Summary:
- Rapid progress in AI is impacting society
- Regulate specific uses and misuses
- Prepare for major economic disruption
- Develop the theory and practice of provably beneficial AI
Richard Dawkins summarizes my thought and ideology in this video. If one is believing in physicalism, the whole spectrum of human life can be replicated by artificial intelligence and other artificial solutions. The majority of philosophers share the idea of physicalism and the research of dualistic options seems to be quite passive.
Is there still space for dualism or should we move to discuss the nature of humanity? Is there any reason for biological human beings after the first artificial generation?
https://futurism.com/videos/richard-dawkins-if-ai-ran-the-world-maybe-it...
I updated the article "AI in Finland" on 18th of October. I have found some new people and organizations around AI, but especially the update is based on these two events:
- AI and philosophy https://www.facebook.com/events/1966942386883337
- Technology, our friend? https://www.facebook.com/events/497861633896682/
Will Knight continues his important series on algorithms biases.
Three articles from MIT Tech Review:
https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/
https://www.technologyreview.com/s/608248/biased-algorithms-are-everywhere-and-no-one-seems-to-care/
https://www.technologyreview.com/s/608986/forget-killer-robotsbias-is-the-real-ai-danger/
Related to the article "Concrete problems in AI Safety" and Bostrom's article "The Ethics of Artificial Intelligence" (the part "Ethics in Machine Learning and Other Domain‐Specific AI Algorithms").
You can find the articles from the conference here http://www.fatml.org/schedule/2017/page/papers-2017
Also articles from earlier years (2016, 2015) and an interesting Resources section. Take a look!
Valpola's comments on AI's future, AI and machine learning approaches, including critique to DeepMind's recent papers:
"I've been working on that for a long time, something like eight years ago we started work on something very similar as they are publishing now. Been there, seen that, doesn't work.”