Paresh Kharya

Eric Schmidt's recent interview at Stanford University got lot of attention due to his comments on Google's WFH policy. He mentioned Google is effectively slipping in AI because of their WFH policy. There were lot of people taking sides in this conversation and bashing WFH vs In-office. I would have expected his interview to start healthy discussion around this topic rather than continuing the long running debate on this. Unfortunately, we live in an environment where everyone need to take sides rather than finding positives on both sides. I have worked both in WFH environment (which everyone is forced to during the Covid lockdown) and in-office environment equally over my career so want to write my thoughts on this topic. I have benefitted immensely during the initial days of my career working closely with team in office and having face to face interaction. It certainly offers easier collaboration and provides opportunity for spontaneous discussion(water cooler/tea time). The thing I hated most about working in office is the commute time and the associated stress it adds to the daily life. If you don’t have a team or leader that understands the necessity to have meetings you might be stuck in unproductive/unwanted meetings while in office.    I enjoy the clear separation between work and personal life offered in the office work environment, where you have the chance to log off after your work. Unless you set clear timing/work space in WFH, it is really difficult to separate the two and end up mixing both. WFH offers better work-life balance since there is no commute time and can plan your day better.   But I personally think Hybrid model offers the best of both world and might be what the future workplace would look like, I doubt if we are ever going back to full time in office model or fully remote Covid model. I certainly enjoy the days I am in office and enjoy collaborative discussion with my team. I plan my week to maximize collaboration while in office and do more of my individual planning work during the days I work from home.    I have benefitted from working in all the models and each offers valid arguments to be the de-facto model for a company. Every company/team should try out the different models and choose what works for them. I have seen successful companies in each of the working categories, it is because of the team and the model they choose which worked for them.   It would be great if we have productive discussions to understand why someone chooses a particular model and better understand why it works for a particular team or a company so we can learn from it.  

7

Big AI Problem and a solution So look at this: https://lnkd.in/gjBb9Yiz Microsoft just introduced VASA-1. It's a new AI model that can turn 1 photo and 1 piece of audio into a fully lifelike human deepfake. The difference between truth and falsehood vanishes. This is so scary and dangerous. I WOULD LIKE TO SAY THAT COURT SHOULD NOT ACCEPT ANY VIDEO OR AUDIO THAT IS NOT SIGNED! So what is the solution? A simple mechanism which is quite a norm in the asymmetric cryptography field is delineated below. : 1. First the Govt of each nation should have mechanism to generate key pair for each of its citizen. 2. So everyone has a public and private key. 3. Actually people generate the key pair and send the public key to govt for signing. 4. The private key is only known to the person concerned and not even the government has a copy. 5. The digital certificate having the public key should have Government as the certifying authority. 6. There should be a public repository by the government where in the digital certificates (assume the public key) of the citizens can be downloaded given the Citizen id (e.g. Aadhaar number in India). : Now let us assume that you have an audio or video interview or anything generated by yourself. In any interaction wherein a digital media file is produced the following must be done by a software: 1. Take an electronic finger print (message digest) of the media file. 2. This message digest will be then be encrypted with private key of the person participating. 3. Let us call this message digest encrypted with private key as - signature of the person. 4. If multiple people participate in the video or audio file - then there will be a signature for each of the persons involved. 5. These multiple signatures should be attached to every media file as meta-data. : Now in court of law whenever a video/audio is presented, the following will happen: 1. The signatures attached to the files are individually decrypted by the person involved using their public key. 2. The finger-print will then be compared to the calculated finger-print of the media file. 3. If it does not match then the person was not involved. Or if it does not decrypt using the public key then too the person is not involved. 4. Remember private key is known only to the person and no one else. : A sincere appeal to the honourable Chief Justice India (#CJI) - Justice Chandrachud, for changing the legal system wrt to deep fake problem of AI. Security experts sitting with legal luminaries should nix this problem ASAP. We have one less motivation for malicious actors to cause havoc in society with deep fakes. Otherwise all hell will break loose. Tomorrow a deep fake can be generated of any popular person saying anything. BTW, a similar mechanism can be used for electronic voting and is literally tamper proof - no one can play the fool. More foolproof than physical voting as well as EVM. #DeepFake #Law #SupremeCourt #CJI #AI #Security #OpenAI #Microsoft

3

2 Comments