History of AI
There is no set definition of artificial intelligence however it is present everywhere we go, from Siri to
autonomous cars. Artificial intelligence was created in 1956 and it has ups and downs which are represented
as winters and summers. There are four steams of AI from the normal AI to machine learning, to deep
learning and generative ai. In machine learning you have the GOFAI where you give a Function (if you have a
headache) and get an output (you need aspirin). In this stream all the rules are in the system, and you have
control over the outputs. The second one is machine learning, where based on the data the function is
determined. Input + output à function. In deep learning you do not have to explain the input because the
feature and extraction process occur within the machine without an external explanation. Finally generative AI
focuses on the generation of content, like chat gpt and audio. There is a risk of hallucination as chat gpt will
just give you an answer based off the information it has so might give you false info. Additionally, biases
occur as generative AI is trained on data, so if your data is biased the output is based on bias data. Generative
Ai is not very good in logical reasoning, and additional ethical concerns include massive carbon footprints,
monopoly of big tech companies and requires a lot of human workers.
AI Law
The increase in the use of computers gave rise to new legal questions and as time passed research moved into
what opportunities do computers offer in the legal realm.
There is model law to create AI, to support people doing legal work today, and legal questions for AI, legal
instruments which have you to do with the introduction of technology to society.
AI and law started with legal expert systems, go to the expert, and talk to them then put it into the system so
the system would provide answers the expert would, rule-based and the influence mechanism which work
together. The problem with this is that in law you have exceptions, therefore they implemented the exceptions
as conditions to the rule making the use of law very difficult.
Then there was a shift into neural network experiments, where sentencing was based on input data like gender
and age, however this increases biases. You need to identify what inputs are relevant so that the neural net
would work based on set outputs and you can tweak. However, the issue is that you cannot explain why. An
example is the plane landing alone, we don’t care why we just care that it works properly. In the 2000s there
was a shift from rule based to data driven. Finally in 2013 deep learning started to work, where you
summarise case law. And two drivers of this were availability and generation of data and increasing computer
power.
, Both deep learning and rule based have input and output data however deep learning correlate input data to
output data, it does not reason it. Thus, the consequences of data analysis is the example of insurance fees:
later payments cause higher costs. If you don’t understand why your input gives you an output, why are you
using it? Deep learning language models, on the other hand, seem to reason, they argue information and seem
to be capable of giving information and understanding it.
Legal Frameworks
Some history behind legal frameworks. Before data protection there was Privacy, so the UN universal
declaration of Human rights and the ECHR article 8. Privacy was seen both from a physical perspective like
house, body, property, and a non-physical privacy like the mind, feeling and thoughts. In 1980 there was the
first international statement for data protection by OECD. In 1981 Council of Europe made the first legally
binding instrument Convention 108, the council of Europe is not a body of EU, it is an international
organization, drafted ECHR. In1995 the data protection directive was created. In the 2000 the Right to Data
Protection became enshrined in the EU’s charter for Fundamental Rights. Finally in 2016 the regulation of the
EU on the protection of natural persons with regard to the processing of personal data and the free movement
of such data
GDPR- general data protection regulation
The GDPR applied to the processing of personal data wholly or partly by automated means.
The personal data is any information which related to a particular person which is an identified or identifiable,
taking into account all the means likely reasonably to be used, must be a natural person.
Case C-582/14, in order to understand whether means are reasonably likely to be used to identify the natural
person, account should be taken of all objective factors taking into consideration the available technology at
the time of the processing and technological developments. The internet service provider can retrieve which
individual is connected to a dynamic IP address. In this case breyer argues that the government had no right to
store his personal data, Ip address, whilst the government argued that it was not personal data. According to
the GDPR IP address does constitute personal data.
This related to pseudonymisation which means the processing of personal data, in a way that the personal data
can no longer be attributed to a specific data subject without the use of additional information, provided that
the additional information is kept separately and is subject to measures to ensure that the personal data is not
identified or identifiable to a natural person. Thus, meaning that it is still personal data which can be a security
measure. Unless explicit consent or legal obligation is given. Companies are not a natural person for data
protection
There is no set definition of artificial intelligence however it is present everywhere we go, from Siri to
autonomous cars. Artificial intelligence was created in 1956 and it has ups and downs which are represented
as winters and summers. There are four steams of AI from the normal AI to machine learning, to deep
learning and generative ai. In machine learning you have the GOFAI where you give a Function (if you have a
headache) and get an output (you need aspirin). In this stream all the rules are in the system, and you have
control over the outputs. The second one is machine learning, where based on the data the function is
determined. Input + output à function. In deep learning you do not have to explain the input because the
feature and extraction process occur within the machine without an external explanation. Finally generative AI
focuses on the generation of content, like chat gpt and audio. There is a risk of hallucination as chat gpt will
just give you an answer based off the information it has so might give you false info. Additionally, biases
occur as generative AI is trained on data, so if your data is biased the output is based on bias data. Generative
Ai is not very good in logical reasoning, and additional ethical concerns include massive carbon footprints,
monopoly of big tech companies and requires a lot of human workers.
AI Law
The increase in the use of computers gave rise to new legal questions and as time passed research moved into
what opportunities do computers offer in the legal realm.
There is model law to create AI, to support people doing legal work today, and legal questions for AI, legal
instruments which have you to do with the introduction of technology to society.
AI and law started with legal expert systems, go to the expert, and talk to them then put it into the system so
the system would provide answers the expert would, rule-based and the influence mechanism which work
together. The problem with this is that in law you have exceptions, therefore they implemented the exceptions
as conditions to the rule making the use of law very difficult.
Then there was a shift into neural network experiments, where sentencing was based on input data like gender
and age, however this increases biases. You need to identify what inputs are relevant so that the neural net
would work based on set outputs and you can tweak. However, the issue is that you cannot explain why. An
example is the plane landing alone, we don’t care why we just care that it works properly. In the 2000s there
was a shift from rule based to data driven. Finally in 2013 deep learning started to work, where you
summarise case law. And two drivers of this were availability and generation of data and increasing computer
power.
, Both deep learning and rule based have input and output data however deep learning correlate input data to
output data, it does not reason it. Thus, the consequences of data analysis is the example of insurance fees:
later payments cause higher costs. If you don’t understand why your input gives you an output, why are you
using it? Deep learning language models, on the other hand, seem to reason, they argue information and seem
to be capable of giving information and understanding it.
Legal Frameworks
Some history behind legal frameworks. Before data protection there was Privacy, so the UN universal
declaration of Human rights and the ECHR article 8. Privacy was seen both from a physical perspective like
house, body, property, and a non-physical privacy like the mind, feeling and thoughts. In 1980 there was the
first international statement for data protection by OECD. In 1981 Council of Europe made the first legally
binding instrument Convention 108, the council of Europe is not a body of EU, it is an international
organization, drafted ECHR. In1995 the data protection directive was created. In the 2000 the Right to Data
Protection became enshrined in the EU’s charter for Fundamental Rights. Finally in 2016 the regulation of the
EU on the protection of natural persons with regard to the processing of personal data and the free movement
of such data
GDPR- general data protection regulation
The GDPR applied to the processing of personal data wholly or partly by automated means.
The personal data is any information which related to a particular person which is an identified or identifiable,
taking into account all the means likely reasonably to be used, must be a natural person.
Case C-582/14, in order to understand whether means are reasonably likely to be used to identify the natural
person, account should be taken of all objective factors taking into consideration the available technology at
the time of the processing and technological developments. The internet service provider can retrieve which
individual is connected to a dynamic IP address. In this case breyer argues that the government had no right to
store his personal data, Ip address, whilst the government argued that it was not personal data. According to
the GDPR IP address does constitute personal data.
This related to pseudonymisation which means the processing of personal data, in a way that the personal data
can no longer be attributed to a specific data subject without the use of additional information, provided that
the additional information is kept separately and is subject to measures to ensure that the personal data is not
identified or identifiable to a natural person. Thus, meaning that it is still personal data which can be a security
measure. Unless explicit consent or legal obligation is given. Companies are not a natural person for data
protection
