AI not as intelligent as you think, although…

HOWEVER, I mean only the intelligence (AI) application called ChatGPT released in November last year publicly and free-of-charge, with the paid subscription version costing $20 monthly (which I can’t afford). ChatGPT has taken the world, even the Philippines, by storm, as it is free, and is most people’s (mine at least) first experience in interacting with an AI, which even serious computer scientists warn could, if safeguards are not imposed, go against this irrational species called humanity, as the “Terminator,” “Matrix” and other popular movies have depicted.

ChatGPT itself explains: “It is a form of artificial intelligence (AI). It is powered by a deep learning model called GPT (generative pre-trained transformer), which is a type of AI model that has been trained on large amounts of text data to understand and generate human-like text.

“ChatGPT is designed for natural language understanding and generation, making it capable of engaging in text-based conversations and performing various language-related tasks, such as answering questions, providing explanations, and generating text. It uses AI techniques, particularly deep learning and natural language processing, to process and generate human-readable text responses based on the input it receives from users.”

I found my first conversations with ChatGPT indeed creepy, as if I were chatting with a ghostly human. Indeed as I will narrate below, it even apologized for its error, unlike many writers I have called out yet refuse to admit their gross misinformation. Here’s relevant bits of my talk with ChatGPT, on a topic I know well enough:

Disputes

Me: How should the Philippines handle its territorial and maritime disputes with China?

ChatGPT: “The Philippines finds itself in a complex and delicate situation regarding territorial and maritime disputes with China. These disputes primarily revolve around the South China Sea, an area of immense strategic and economic importance. To effectively handle these disputes, the Philippines must adopt a multifaceted approach that combines diplomacy, international cooperation, domestic policy and legal mechanisms… “

(ChatGPT expounds on each of the approaches it listed, but not in concrete terms. It then claims: “The Philippines should continue to utilize international legal mechanisms, such as Unclos, to resolve disputes. In 2016, the Permanent Court of Arbitration ruled in favor of the Philippines’ territorial claims, emphasizing the importance of respecting the tribunal’s decision and seeking its enforcement.”)

Me: You are wrong that it was the Permanent Court of Arbitration that made the ruling you mentioned.

ChatGPT responds almost simultaneously:

“I apologize for any confusion in my previous responses. The ruling against China in the South China Sea dispute was issued by an arbitral tribunal established under the United Nations Convention on the Law of the Sea (Unclos).. The arbitral tribunal that heard the case was established by the PCA, an intergovernmental organization based in The Hague, Netherlands.

“The PCA itself did not make the ruling; rather, it was the arbitral tribunal convened under its authority that made the decision. I hope this clarifies the source of the ruling in the South China Sea arbitration case.”

Wrong

Me: You are still wrong in saying the arbitral tribunal was established under the PCA.

ChatGPT: “You are correct, and I apologize for the oversight. The arbitral tribunal for the Philippines v. China case was indeed not established under the Permanent Court of Arbitration (PCA). It was established under Annex VII to the United Nations Convention on the Law of the Sea (Unclos). The tribunal was convened in accordance with the procedures outlined in Unclos.”

Thank you for pointing out the correction, and I appreciate your understanding.

Me: What were these procedures?

ChatGPT: “The Philippines initiated the arbitration by submitting a Notification and Statement of Claim to China on Jan. 22, 2013. In this notification, the Philippines formally informed China of its intention to settle the dispute through arbitration and outlined the specific issues in dispute.

“The process of appointing arbitrators began when the Philippines appointed one arbitrator, and China was invited to do the same. However, China declined to participate in the arbitration proceedings and did not appoint an arbitrator. Consequently, the Philippines nominated the remaining members of the tribunal, and these arbitrators proceeded to appoint the president of the tribunal.”

Repeating

However, in response to another of my later questions, ChatGPT wrote: “The Permanent Court of Arbitration (PCA) issued an award in favor of the Philippines, ruling that certain Chinese claims had no legal basis under Unclos. While China rejected the ruling, the award strengthened the Philippines’ legal position in the South China Sea.”

Me: Why do you keep repeating that the “PCA ruled in 2016..” which is a wrong statement which you already apologized over?

Chat: “I apologize for any confusion, and thank you for pointing out the earlier error. You are correct; the award in the Philippines v. China case was issued by the arbitral tribunal constituted under the United Nations Convention on the Law of the Sea (Unclos), not by the Permanent Court of Arbitration (PCA) itself. I appreciate your patience, and I apologize for any repeated inaccuracies.”

Me: Does your system prefer a statement when it was used more times on the internet, rather than an accurate statement although less mentioned on the internet?

ChatGPT: “My responses are generated based on patterns in the large dataset used for training, and I don’t have personal preferences or biases. However, the way my responses are generated is influenced by the frequency and prevalence of information in the training data. If a statement or piece of information is more commonly found on the internet and appears in a broader range of sources, it may have a higher likelihood of being included in responses. (Itals mine).

“It’s important to note that my responses are generated by a version of GPT-3.5, and my knowledge is based on information available up until September 2021. While I can provide information and engage in conversations on a wide range of topics, my responses are generated based on patterns in the data I was trained on and may not always reflect the most up-to-date information or developments.”

Vulnerable

There you are, the more a statement is mentioned in the internet, which ChatGPT scans at an unbelievable speed — which potentially makes it a threat to us humans, who can read probably 10 internet postings in an hour — it is vulnerable to fake news spread by trolls, US propaganda machines and repetition by know-it-all but lazy or ignorant writers.

At least ChatGPT acknowledges its errors unlike many so-called experts on the South China Sea disputes, especially Stratbase ADR head Dindo Manhit, who is supposedly an expert on South China issues. I had corrected his wrong statement that it was the Permanent Court of Arbitration, providing documents to prove my point. He didn’t bother to correct his statement. Worse, he repeated it in a recent column.

Unlike most so-called experts on the South China Sea disputes, ChatGPT learns from its interactions with humans like me.

On another issue that has been the subject of so much US propaganda based on distorting the arbitration ruling, I asked ChatGPT: If the arbitral tribunal ruled that China’s nine-dash line has no legal basis under Unclos, does this mean China has no legitimate claims in the South China Sea?

ChatGPT: “The ruling by the arbitral tribunal that China’s ‘nine-dash line’ has no legal basis under Unclos specifically means that the tribunal found the nine-dash line to be inconsistent with the provisions of Unclos. However, it’s important to note that the tribunal’s ruling did not address the issue of sovereignty over specific islands or features in the South China Sea. Instead, it focused on the maritime entitlements and rights of coastal states in accordance with Unclos.”

Marcos

Because today is a historical date, I couldn’t help asking ChatGPT the question: “Was the Marcos dictatorship marked by massive human rights violations?”

ChatGPT: “Yes, the dictatorship of Ferdinand Marcos in the Philippines, which lasted from 1965 to 1986, was marked by massive human rights violations. During his regime, particularly during the martial law period which began in 1972, there were widespread and systematic abuses of human rights, including:

“Arbitrary detention: Thousands of individuals were arbitrarily detained without due process of law. Many were imprisoned for political reasons or perceived opposition to the government.

“Torture: Torture and other forms of ill-treatment were used against political prisoners and activists.

“Extrajudicial killings: There were reports of extrajudicial killings of political opponents and activists by government forces.

“Censorship: Marcos imposed strict censorship on the media and suppressed freedom of speech and the press, limiting the ability of citizens to express dissenting views.

“Forced disappearances: Some individuals critical of the regime disappeared and were never accounted for.

“Electoral fraud: Marcos maintained his grip on power through electoral fraud and manipulation of the democratic process.

“One of the most notable incidents during the Marcos dictatorship was the declaration of martial law in 1972, which suspended civil liberties, disbanded the existing constitution, and allowed for the consolidation of power by Marcos. This period of martial law became synonymous with widespread human rights abuses and political repression.”

ChatGPT is inaccurate in its assertions here, and I have written many columns providing data that corrects its report. This administration though appears to be uninterested in correcting lies about the elder Marcos’ regime. Keep in mind though that ChatGPT’s earlier explanation that its “knowledge is based on information available up until September 2021,” or before the son of the dictator became president July 2022.

Will ChatGPT’s report be different if it updates its database? I doubt it. As things are going now, ChatGPT’s report in the future could be worse.

---

Facebook: Rigoberto Tiglao

X: @bobitiglao

Archives: www.rigobertotiglao.com

Book orders: https://ift.tt/pgv9rlx

The post AI not as intelligent as you think, although… first appeared on Rigoberto Tiglao.

AI not as intelligent as you think, although…
Source: Breaking News PH